Updated June 26, 2003
Publications: 2003, 2002, 2001,...
Syllabus: CN 550 (Spring 2002) -- Neural and Computational Models of Recognition, Memory, and Attention
Download: http://cns.bu.edu/~gail/CN 550 Syllabus 2002 .pdf
Department of Cognitive and Neural Systems
Boston University
677 Beacon Street
Boston, Massachusetts 02215
Office: (617) 353-9483
Department office: (617) 353-9481
Fax: (617) 353-7755
gail@bu.edu
Neural networks, learning and memory, attention, pattern recognition, remote sensing, medical database analysis, machine learning, computer vision, synaptic transmission, dynamical systems, differential equations
University of Wisconsin, Madison
Department of Mathematics Ph.D. 1974, M.A. 1972
Thesis: "Traveling wave solutions of nerve impulse equations"
NSF Graduate Fellowship, 1970-1973
Mathematics Research Center Fellowship, 1973-1974
University of Colorado, Boulder
B.A. 1970 (summa cum laude, mathematics)
Massachusetts Institute of Technology (1974-1976)
Instructor in Applied Mathematics
Northeastern University, Department of Mathematics (1976-1989)
Assistant Professor (1976-1979)
Associate Professor (1979-1986)
Professor (1986-1989)
Boston University (1982 - )
Research Associate, Center for Adaptive Systems (1982-1989)
Professor of Cognitive Neural Systems and Mathematics (1989- )
Co-Director, Cognitive Neural Systems Program (1989-1991)
Director of Graduate Studies, Department of Cognitive Neural Systems (1991- )
Director of the CNS Technology Lab (2002- )
Gabor Award (1999), International Neural Network Society
Slovak Artificial Intelligence Society Award (2002)
International Neural Network Society
Governing Board (1987 - )
Secretary (1994 - 2000)
Executive Committee (1994 - 2000)
Vice President (1988-1989)
American Mathematical Society (1996-1999)
AMS Council Member-at-Large
Committee on the Profession
Brain Research (Cognitive Brain Research and Computational Neuroscience Section)
Encyclopedia of Cognitive Science
IEEE Transactions on Neural Networks
Neural Computation
Neural Networks
Neural Processing Letters
Computer Abstracts
American Mathematical Society (AMS)
Association for Women in Mathematics (AWM)
Institute of Electrical and Electronics Engineers (IEEE)
International Neural Network Society (INNS)
Society for Industrial Applied Mathematics (SIAM)
Society for Neuroscience (SN)
1. Carpenter, G.A. (Editor) (1987). Some Mathematical Questions in Biology: Circadian Rhythms, Lectures on Mathematics in the Life Sciences, Volume 19, American Mathematical Society.
2. Carpenter, G.A. & Grossberg, S. (Editors) (1987). Applied Optics : Special issue on neural networks, 26(23).
3. Carpenter, G.A. & Grossberg, S. (Editors) (1991). Pattern Recognition by Self-Organizing Neural Networks, Cambridge, MA: MIT Press.
4. Carpenter, G.A. & Grossberg,
S. (Editors) (1992). Neural Networks for Vision and Image Processing,
Cambridge, MA: MIT Press.
1. Carpenter, G.A. & Grossberg, S., U.S. Patent No. 5,142,590: Pattern recognition system (Filed: November 27, 1985. Issued: August 25, 1992. European Patent No. 0244483, issued July 15, 1992). Based on article 26 (ART 1).
2. Carpenter, G.A. & Grossberg, S., U.S. Patent Nos. 4,914,708 and 5,133,021: System for self-or of stable category recognition codes for analog patterns (Filed: June 19, 1987. Issued: April 3, 1990 and July 21, 1992). Based on article 31 (ART 2).
3. Carpenter, G.A. & Grossberg, S., U.S. Patent No. 5,311,601: Pattern recognition system with variable selection weights (Filed: January 12, 1990. Issued: May 10, 1994). Based on article 42 (ART 3).
4. Carpenter, G.A., Grossberg, S., & Rosen, D.B., US Patent No. 5,157,738: Rapid category learning and recognition system (Filed: December 19, 1990. Issued: October 20, 1992). Based on article 53 (ART 2- A).
5. Carpenter, G.A., Grossberg, S., & Reynolds, J.H., US Patent No. 5,214,715: Predictive self-organizing neural network. (Filed: January 31, 1991. Issued: May 25, 1993). Based on article 54 (ARTMAP).
1. Carpenter, G.A. (1976). Nerve impulse equations. In P. Hilton (Ed.), Structural Stability, The Theory of Catastrophes, and Applications in the Sciences, Springer Series: Lecture Notes in Mathematics, 525, 58-76.
2. Carpenter, G.A. (1976). Models of excitable membrane phenomena. Brain Theory Newsletter, March, 42-44.
ARTICLES (1977)
3. Carpenter, G.A. (1977). A geometric approach to singular perturbation problems with applications to nerve impulse equations. Journal of Differential Equations, 23, 335-367.
Download Hodgkin-Huxley (1952) paper summary: http://www.jphysiol.org/cgi/reprint/538/1/2
4. Carpenter, G.A. (1977). Periodic solutions of nerve impulse equations. Journal of Mathematical Analysis and Applications, 58, 152-173.
ARTICLES (1978)
5. Carpenter, G.A. (1978). A mathematical analysis of excitable membrane phenomena. In R. Trappl, G.J. Klir, & L. Ricciardi (Eds.), Progress in Cybernetics and Systems Research: General Systems Methodology, Fuzzy Mathematics and Fuzzy Systems, Biocybernetics and Theoretical Neurobiology, New York: Halsted Press, 3, 58- 76.
6. Carpenter, G.A. & Knapp, V. (1978). An analysis of the mammalian ventricular action potential. Journal of Mathematical Biology, 6, 305-316.
ARTICLES (1979)
7. Carpenter, G.A. (1979). Bursting phenomena in excitable membranes. SIAM Journal on Applied Mathematics, 36, 334-372.
ARTICLES (1981)
8. Carpenter, G.A. (1981). Normal and abnormal signal patterns in nerve cells. In S. Grossberg (Ed.), Mathematical Psychology and Psychophysiology, AMS/SIAM Symposium Series, 13, 49-90.
9. Carpenter, G.A. & Grossberg, S. (1981). Adaptation and transmitter gating in vertebrate photoreceptors. Journal of Theoretical Neurobiology, 1, 1-42.
Reprinted in: S. Grossberg (Ed.) (1987) The Adaptive Brain. Amsterdam: North Holland.
ARTICLES (1983)
10. Ayers, J.L., Carpenter, G.A., Currie, S., & Kinch, J. (1983). Which behavior does the lamprey central motor program mediate? Science, 221, 1312-1314.
11. Carpenter, G.A. & Grossberg, S. (1983). Dynamic models of neural systems: Propagated signals, photoreceptor transduction, and circadian rhythms. In J.P.E. Hodgson (Ed.), Oscillations in Mathematical Biology, Springer Series: Lecture Notes in Biology, 51, 102-196.
12. Carpenter, G.A. & Grossberg, S. (1983). A neural theory of circadian rhythms: The gated pacemaker. Biological Cybernetics, 48, 35-59.
Reprinted in: S. Grossberg (Ed.) (1987) The Adaptive Brain, Amsterdam: North Holland.
13. Carpenter, G.A. (1983). A comparative analysis of structure and chaos in models of single nerve cells and circadian rhythms. In E. Basar, H. Flohr, H. Haken, & A.J. Mandell (Eds.), Synergetics of the Brain, Springer Series in Synergetics, 23, 311-329.
ARTICLES (1984)
14. Carpenter, G.A. (1984). Gated dipoles: Photoreceptors and circadian rhythms. Behavioral Processes, 9, 90- 91.
15. Carpenter, G.A. & Grossberg, S. (1984). A neural theory of circadian rhythms: Aschoff's rule in diurnal and nocturnal mammals. American Journal of Physiology (Regulatory, Integrative and Comparative Physiology), 247, R1067-R1082.
Reprinted in: S. Grossberg (Ed.) (1987) The Adaptive Brain, Amsterdam: North Holland.
ARTICLES (1985)
16. Carpenter, G.A. & Grossberg, S. (1985). Neural dynamics of circadian rhythms: The hypothalamic pacemaker. In J. Eisenfeld & C. DeLisi (Eds.), Mathematics and Computers in Biomedical Applications, Elsevier Science Publishers (North-Holland), 79-102.
17. Carpenter, G.A. & Grossberg, S. (1985). A neural theory of circadian rhythms: Split rhythms, after-effects, and motivational interactions. Journal of Theoretical Biology, 113, 163-223.
Reprinted in: S. Grossberg (Ed.) (1987) The Adaptive Brain, Amsterdam: North Holland.
18. Carpenter, G.A. (1985). The non-parametric influence of light on mammalian circadian rhythms. In B.D. Sleeman & R.J. Jarvis (Eds.), Ordinary and Partial Differential Equations, Springer Series: Lecture Notes in Mathematics, 1151, 90-108.
19. Carpenter, G.A. (1985). The circadian activity rhythm of mammals: A comparison of models and experiments. In L. Rensing & N.I. Jaeger (Eds.), Temporal Order: Oscillations in Heterogeneous Chemical and Biological Systems, Springer Series in Synergetics, 29, 263-272.
20. Carpenter, G.A. & Grossberg, S. (1985). Category learning and adaptive pattern recognition: A neural network model. Proceedings of the 3rd Army Conference on Applied Mathematics and Computing, ARO Report 86-1, 37- 56.
ARTICLES (1986)
21. Carpenter, G.A. (1986). Some cell-biological mechanisms governing complex neural signal patterns. In S. Diner, D. Farque, & G. Lochak (Eds.), Dynamical Systems A Renewal of Mechanism, Singapore/Philadelphia: World Scientific Publishing Co., 180-200.
22. Carpenter, G.A. (1986). Book review: Mathematical Models of the Circadian Sleep-Wake Cycle, Martin C. Moore-Ede & Charles A. Czeisler (Eds.), Mathematical Biosciences, 79, 231-233.
23. Carpenter, G.A. & Grossberg, S. (1986). Absolutely stable learning of recognition codes by a self-organizing neural network. In J.S. Denker (Ed.), Neural Networks for Computing, American Institute of Physics, 151, 77- 85.
ARTICLES (1987)
24. Carpenter, G.A. & Grossberg, S. (1987). Discovering order in chaos: Stable self-organization of neural recognition codes. In S.H. Koslow, A.J. Mandell, & M.F. Shlesinger (Eds.), Perspectives in Biological Dynamics and Theoretical Medicine, Annals of the New York Academy of Sciences, 504, 33-51.
25. Carpenter, G.A. & Grossberg, S. (1987). Mammalian circadian rhythms: A neural network model. In G.A. Carpenter (Ed.), Some Mathematical Questions in Biology Circadian Rhythms, American Mathematical Society, Lectures on Mathematics in the Life Sciences, 19, 151-203.
26. Carpenter, G.A. & Grossberg, S. (1987). A massively parallel architecture for a self-organizing neural pattern recognition machine. Computer Vision, Graphics, and Image Processing, 37, 54-115.
Reprinted in: C. Lau (Ed.) (1992) Neural Networks: Theoretical Foundations, Piscataway, NJ: IEEE Press; G.A. Carpenter & S. Grossberg (Eds.) (1991) Pattern Recognition by Self-Organizing Neural Networks, Cambridge, MA: MIT Press; and S. Grossberg (Ed.) (1988) Neural Networks and Natural Intelligence, Cambridge, MA: MIT Press.
27. Carpenter, G.A., Cohen, M.A., & Grossberg, S. (1987). Computing with neural networks: The role of symmetry. Science, 235, 1226-1227.
28. Carpenter, G.A. & Grossberg, S. (1987). Associative learning, adaptive pattern recognition, and cooperative- competitive decision making. In H. Szu (Ed.), Optical and Hybrid Computing, SPIE, 634, 218-247.
29. Carpenter, G.A. & Grossberg, S. (1987). ART 2: Self-organization of stable category recognition codes for analog input patterns. Proceedings of the IEEE First International Conference on Neural Networks, II, 727-735.
30. Carpenter, G.A. & Grossberg, S. (1987). Invariant pattern recognition and recall by an attentive ART architecture in a nonstationary world. Proceedings of the IEEE First International Conference on Neural Networks, II, 737-745.
31. Carpenter, G.A. & Grossberg, S. (1987). ART 2: Self-organization of stable category recognition codes for analog input patterns. Applied Optics : Special Issue on Neural Networks, 26, 4919-4930.
Reprinted in: G.A. Carpenter & S. Grossberg (Eds.) (1991) Pattern Recognition by Self-Organizing Neural Networks, Cambridge, MA: MIT Press; and E. Rosenfeld, J.A. Anderson, & A. Pellionisz (Eds.) (1990) Neurocomputing 2: Directions for Research, Cambridge, MA: MIT Press.
ARTICLES (1988)
32. Carpenter, G.A. & Grossberg, S. (1988). Adaptive resonance theory: Stable self-organization of neural recognition codes in response to arbitrary lists of input patterns. Proceedings of the 8th Conference of the Cognitive Science Society, Hillsdale, NJ: Erlbaum Associates, 45-62.
33. Carpenter, G.A. & Grossberg, S. (1988). Neural dynamics of category learning and recognition: Attention, memory consolidation, and amnesia. In J.L. Davis, R.W. Newburgh, &E.J. Wegman (Eds.), Brain Structure, Learning, and Memory, AAAS Selected Symposia Series, 105, 233-290.
Reprinted in: S. Grossberg (Ed.) (1987) The Adaptive Brain, Amsterdam: North-Holland.
34. Carpenter, G.A. & Grossberg, S. (1988). The ART of adaptive pattern recognition by a self-organizing neural network. Computer : Special Issue on Artificial Neural Systems, 21, 77-88.
Reprinted in: V. Vemuri & F.U. Dowla, (Eds.) (1991) Neural Networks Applications in Signal Processing, Image Understanding, and Optimization, Los Alamitos, CA: IEEE Computer Society Press; Pankaj Mehra & Benjamin W. Wah (Eds.) (1992) Artificial Neural Networks: Concepts and Theory, Los Alamitos, CA: IEEE Computer Society Press; and J. Diederich (Ed.) (1990) Concept Learning: Artificial Neural Networks, Los Alamitos, CA: IEEE Computer Society Press.
35. Carpenter, G.A. (1988). Self-organizing neural network architectures for real-time adaptive pattern recognition. In H. Haken (Ed.), Neural and Synergetic Computers, Springer Series in Synergetics, 42, 42- 74.
ARTICLES (1989)
36. Carpenter G.A. & Grossberg, S. (1989). Search mechanisms for Adaptive Resonance Theory (ART) architectures. Proceedings of the IEEE/INNS International Joint Conference on Neural Networks, I, 201- 205.
37. Carpenter, G.A., Grossberg, S., & Mehanian, C. (1989). Invariant recognition of cluttered scenes by a self- organizing ART architecture: CORT-X boundary segmentation. Neural Networks, 2, 169-181.
Download: http://cns.bu.edu/~gail/037_CORT-X_1989_.pdf
38. Carpenter, G.A. (1989). Neural network models for pattern recognition and associative memory. Neural Networks, 2, 243-257.
Reprinted in:
V. Rao Vemuri (Ed.) (1992) Artificial Neural Networks: Concepts and Control Applications, Los Alamitos, CA: IEEE Computer Society Press; Lynn Nadel & Daniel Stein (Eds.) (1992) 1991 Lectures in Complex Systems, Santa Fe Institute Studies in the Sciences of Complexity, Lectures Vol. IV, Redwood City, CA: Addison-Wesley; V. Vemuri & F.U. Dowla (Eds.) (1991) Neural Networks Applications in Signal Processing, Image Understanding, and Optimization, Los Alamitos, CA: IEEE Computer Society Press; G.A. Carpenter & S. Grossberg (Eds.) (1991) Pattern Recognition by Self-Organizing Neural Networks, Cambridge, MA: MIT Press; and Edgar Sanchez-Sinencio & Clifford Lau (Eds.) (1991) Neural Networks in Circuits and Systems, Piscataway, NJ: IEEE Press.
Download: http://cns.bu.edu/~gail/038_NN_Review_1989_.pdf
ARTICLES (1990)
39. Carpenter, G.A. & Grossberg, S. (1990). Neural dynamics of category learning and recognition: Structural invariants, evoked potentials, and reinforcement. In M.L. Commons, R.J. Herrnstein, S.M. Kosslyn, & D.B. Mumford (Eds.), Quantitative Analyses of Behavior, IX: Computational and Clinical Approaches to Pattern Recognition and Concept Formation, Hillsdale, NJ: Erlbaum Associates, 23-49.
40. Carpenter, G.A. & Grossberg, S. (1990). ART 3 hierarchical search: Chemical transmitters in self-organizing pattern recognition architectures. Proceedings of the INNS/IEEE International Joint Conference on Neural Networks, II, Hillsdale, NJ: Erlbaum Associates, 30-33.
41. Carpenter, G.A. & Grossberg, S. (1990). Adaptive resonance theory: Neural network architectures for self- organizing pattern recognition. In R. Eckmiller, G. Hartmann, & G. Hauske (Eds.), Parallel Processing in Neural Systems and Computers, Amsterdam: North Holland, 383-389.
42. Carpenter, G.A. & Grossberg, S. (1990). ART 3: Hierarchical search using chemical transmitters in self- organizing pattern recognition architectures. Neural Networks, 3, 129-152.
Reprinted in: Ray Paton (Ed.) (1994) Computing with Biological Metaphors, London: Chapman and Hall Ltd.; Edgar Sanchez-Sinencio & Clifford Lau (Eds.) (1991) Neural Networks in Circuits and Systems, Piscataway, NJ: IEEE Press; and G.A. Carpenter & S. Grossberg (Eds.) (1991) Pattern Recognition by Self-Organizing Neural Networks, Cambridge, MA: MIT Press.
Download: http://cns.bu.edu/~gail/042_ART_3_1990_.pdf
43. Carpenter, G.A. & Grossberg, S. (1990). ART 3: Self-organization of distributed pattern recognition codes in neural network hierarchies. Proceedings of the International Neural Network Conference (Paris), Dordrecht: Kluwer Academic Publishing, 801-804.
44. Carpenter, G.A. & Grossberg, S. (1990). Self-organizing neural network architectures for real-time adaptive pattern recognition. In S.F. Zornetzer, J.L. Davis, & C. Lau (Eds.), An Introduction to Neural and Electronic Networks, San Diego: Academic Press, 455-478.
45. Carpenter, G.A. & Grossberg, S. (1990). ART: Self-organizing neural networks for learning and memory of cognitive recognition codes. Proceedings of the 12th Annual Conference of the Cognitive Science Society, Hillsdale, NJ: Erlbaum Associates, 1032-1034.
ARTICLES (1991)
46. Bullock, D., Carpenter, G.A., & Grossberg, S. (1991). Self-organizing neural network architectures for adaptive pattern recognition and robotics. In P. Antognetti & V. Milutinovic (Eds.), Neural Networks: Concepts, Applications, and Implementations, I, Englewood Cliffs, NJ: Prentice-Hall, 33-53.
47. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1991). A self-organizing ARTMAP neural architecture for supervised learning and pattern recognition. In R. J. Mammone & Y.Y. Zeevi (Eds.), Neural Networks: Theory and Applications, New York: Academic Press, 43-80.
48. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1991). A self-organizing ARTMAP neural architecture for supervised learning and pattern recognition. In T. Kohonen, K. Mäkasira, O. Simula, & J. Kangas (Eds.), Artificial Neural Networks, Amsterdam: North-Holland/Elsevier Science Publishing, I-31-36.
49. Bradski, G., Carpenter, G.A., & Grossberg, S. (1991). Working memory networks for learning multiple groupings of temporally ordered events: Applications to 3-D visual object recognition. Proceedings of the International Joint Conference on Neural Networks (IJCNN-91), Piscataway, NJ: IEEE Service Center, I- 723-728. Technical Report CAS/CNS-TR-91-007, Boston, MA: Boston University.
50. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1991). ARTMAP: A self-organizing neural network architecture for fast supervised learning and pattern recognition. Proceedings of the International Joint Conference on Neural Networks (IJCNN-91), Piscataway, NJ: IEEE Service Center, I-863-868.
51. Carpenter, G.A., Grossberg, S., & Rosen, D.B. (1991). ART 2-A: An adaptive resonance algorithm for rapid category learning and recognition. Proceedings of the International Joint Conference on Neural Networks (IJCNN- 91), Piscataway, NJ: IEEE Service Center, II-151-156.
52. Carpenter, G.A., Grossberg, S., & Rosen, D.B. (1991). Fuzzy ART: An adaptive resonance algorithm for rapid, stable classification of analog patterns. Proceedings of the International Joint Conference on Neural Networks (IJCNN-91), Piscataway, NJ: IEEE Service Center, II-411-416. Technical Report CAS/CNS-TR- 91-006, Boston, MA: Boston University.
53. Carpenter, G.A., Grossberg, S., & Rosen, D.B. (1991). ART 2-A: An adaptive resonance algorithm for rapid category learning and recognition. Neural Networks, 4, 493-504. Technical Report CAS/CNS-TR-91-011, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/053_ART_2-A_1991_.pdf
54. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1991). ARTMAP: Supervised real-time learning and classification of nonstationary data by a self-organizing neural network. Neural Networks, 4, 565-588. Technical Report CAS/CNS-TR-91-001, Boston, MA: Boston University.
Reprinted in: G.A. Carpenter & S. Grossberg (Eds.) (1991) Pattern Recognition by Self-Organizing Neural Networks, Cambridge, MA: MIT Press.
Download: http://cns.bu.edu/~gail/054_ARTMAP_1991_.pdf
55. Carpenter, G.A. & Grossberg, S. (1991). Distributed hypothesis testing, attention shifts, and transmitter dynamics during the self-organization of brain recognition codes. In H.G. Schuster & W. Singer (Eds.), Nonlinear Dynamics and Neuronal Networks, New York: Springer-Verlag, 305-334. Technical Report CAS/CNS-TR-91-013, Boston, MA: Boston University.
56. Carpenter, G.A., Grossberg, S., & Rosen, D.B. (1991). Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system. Neural Networks, 4, 759-771. Technical Report CAS/CNS- TR-91-015, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/056_Fuzzy_ART_1991_.pdf
57. Carpenter, G.A., Grossberg, S., & Rosen, D.B. (1991). A neural network realization of fuzzy ART. Technical Report CAS/CNS-TR-91-021, Boston, MA: Boston University.
58. Carpenter, G.A. & Grossberg, S. (1991). Attention, resonance, and transmitter dynamics in models of self- organizing cortical networks for recognition learning. In A.V. Holden & V.I. Kryukov (Eds.), Neurocomputers and Attention, Volume I: Neurobiology, Synchronization and Chaos, Manchester: Manchester University Press, 201-222.
ARTICLES (1992)
59. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1992). A neural network architecture for fast on-line supervised learning and pattern recognition. In H. Wechsler (Ed.), Neural Networks for Perception. Volume 1: Human and Machine Perception, New York: Academic Press, 248-264.
60. Carpenter, G.A. & Grossberg, S. (1992). Self-organizing cortical networks for distributed hypothesis testing and recognition learning. In J.G. Taylor & C.L.T. Mannion (Eds.), Theory and Applications of Neural Networks, New York: Springer-Verlag, 3-27.
61. Carpenter, G.A. & Grossberg, S. (1992). Adaptive resonance theory. In Stuart C. Shapiro (Ed.), Encyclopedia of Artificial Intelligence, Second Edition. New York: Wiley and Sons, 13-21.
62. Bradski, G., Carpenter, G.A., & Grossberg, S. (1992). Working memory networks for learning temporal order, with application to 3-D visual object recognition. Neural Computation, 4, 270-286. Technical Report CAS/CNS-TR-91-014, Boston, MA: Boston University.
63. Carpenter, G.A., Grossberg, S. Markuzon, N. Reynolds, J.H. & Rosen, D.B. (1992). Attentive supervised learning and recognition by an adaptive resonance system. In G.A. Carpenter & S. Grossberg (Eds.), Neural Networks for Vision and Image Processing, Cambridge, MA: MIT Press, 365-384.
64. Goodman, P.H., Kaburlasos, V.G., Egbert, D.D., Carpenter, G.A., Grossberg, S., Reynolds, J., Hammermeister, K.E., Marshall, G., & Grover, F.L. (1992). Fuzzy ARTMAP neural network prediction of heart surgery mortality. Wang Institute Conference on Neural Networks, 48.
65. Bradski, G., Carpenter, G.A., & Grossberg, S. (1992). Working memories for storage and recall of arbitrary temporal sequences. Proceedings of the International Joint Conference on Neural Networks (IJCNN-92), II-57- 62. Technical Report CAS/CNS-TR-92-003, Boston, MA: Boston University.
66. Carpenter, G.A., Grossberg, S., & Iizuka, K. (1992). Comparative performance measures of Fuzzy ARTMAP, Learned Vector Quantization, and back propagation for handwritten character recognition. Proceedings of the International Joint Conference on Neural Networks (IJCNN-92), I-794-799. Technical Report CAS/CNS- TR-92-005, Boston, MA: Boston University.
67. Carpenter, G.A., Grossberg, S., & Lesher, G. W. (1992). A what-and-where neural network for invariant image processing. Proceedings of the International Joint Conference on Neural Networks (IJCNN-92), III-303- 308. Technical Report CAS/CNS-TR-92-006, Boston, MA: Boston University.
68. Carpenter, G.A., Grossberg, S., Markuzon, N., Reynolds, J.H., & Rosen, D.B. (1992). Fuzzy ARTMAP: An adaptive resonance architecture for incremental learning of analog maps. Proceedings of the International Joint Conference on Neural Networks (IJCNN-92), III-309-314.
69. Carpenter, G.A. & Grossberg, S. (1992). A self-organizing neural network for supervised learning, recognition, and prediction. IEEE Communications Magazine, 30(September), 38-49.
70. Carpenter, G.A., Grossberg, S., Markuzon, N., Reynolds, J.H., & Rosen, D.B. (1992). Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps. IEEE Transactions on Neural Networks, 3, 698-713. Technical Report CAS/CNS-TR-91-016, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/070_Fuzzy_ARTMAP_1992_.pdf
Translated into Japanese and reprinted in: T. Yamakawa (Ed.) (1995) Fuzzy Neural Systems, 12, Japan Society for Fuzzy Theory and Systems, 65-113.
71. Carpenter, G.A. (1992). Book review: Introduction to Neural and Cognitive Modeling, by Daniel S. Levine. IEEE Transactions on Neural Networks, 3, 1030-1031.
72. Goodman, P.H., Kaburlasos, V.G., Egbert, D.D., Carpenter, G.A., Grossberg, S., Reynolds, J.H., Rosen, D.B., & Hartz, A.J. (1992). Fuzzy ARTMAP neural network compared to linear discriminant analysis prediction of the length of hospital stay in patients with pneumonia. Proceedings of the IEEE International Conference on Systems, Man, and Cybernetics. (Chicago, October, 1992) I, New York: IEEE Press, 748-753.
Reprinted in: R.J. Marks, II (Ed.) (1994) Fuzzy Logic Technology and Applications, Piscataway, NJ: IEEE Press.
73. Carpenter, G.A. & Grossberg, S. (1993). Normal and amnesic learning, recognition, and memory by a neural model of cortico-hippocampal interactions. Trends in Neuroscience, 16(4), 131-137. Technical Report CAS/CNS-TR-92-021, Boston, MA: Boston University.
74. Asfour, Y.R., Carpenter, G.A., Grossberg, S. & Lesher, G.W. (1993). Fusion ART A neural network architecture for multi-channel data fusion and classification. Proceedings of the World Congress on Neural Networks (WCNN-93), II-210-215. Technical Report CAS/CNS-TR-93-006, Boston, MA: Boston University.
75. Carpenter, G.A. (1993). Distributed outstar learning and the rules of synaptic transmission. Proceedings of the World Congress on Neural Networks (WCNN-93), II-397-404.
76. Carpenter, G.A. & Govindarajan, K.K. (1993). Evaluation of speaker normalization methods for vowel recognition using fuzzy ARTMAP and K-NN. Proceedings ofthe World Congress on Neural Networks (WCNN-93), III-397-404. Technical Report CAS/CNS-TR-93-013, Boston, MA: Boston University.
77. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1993). Fuzzy ARTMAP, slow learning, and probability estimation. Proceedings of the World Congress on Neural Networks (WCNN-93), II-26-30. Technical Report CAS/CNS-TR-93-014, Boston, MA: Boston University.
78. Carpenter, G.A. & Ross, W.D. (1993). ART-EMAP: A neural network architecture for learning and prediction by evidence accumulation. Proceedings of the World Congress on Neural Networks, (WCNN- 93), III-649-656. Technical Report CAS/CNS-TR-93-015, Boston, MA: Boston University.
79. Carpenter, G.A. & Tan, A.H. (1993). Rule extraction, fuzzy ARTMAP, and medical databases. Proceedings of the World Congress on Neural Networks (WCNN-93), I-501-506. Technical Report CAS/CNS-TR-93- 016, Boston, MA: Boston University.
80. Carpenter, G.A., Grossberg, S., Markuzon, N., Reynolds, J.H., & Rosen, D.B. (1993). Supervised learning by adaptive resonance neural networks. In M. Marinaro & G. Scarpetta (Eds.), Structure: From Physics to General Systems, 2. Festschrift volume in honor of the 70th birthday of Professor Eduardo R. Caianiello. Singapore: World Scientific Publishing Co., 36-47.
81. Carpenter, G.A. & Govindarajan, K.K. (1993). Neural network and nearest neighbor comparison of speaker normalization methods for vowel recognition. In S. Gielen & B. Kappen (Eds.), Proceedings of the International Conference on Artificial Neural Networks (ICANN'93), London, UK: Springer-Verlag, 412-415.
82. Asfour, Y.R., Carpenter, G.A., Grossberg, S. & Lesher, G. W. (1993). Fusion ART An adaptive fuzzy network for multi-channel classification. Proceedings of the Third International Conference on Industrial Fuzzy Control and Intelligent Systems (IFIS-93, Houston), Piscataway, NJ: IEEE Press, 155-160. Technical Report CAS/CNS TR-93-052, Boston, MA: Boston University.
83. Carpenter, G.A. (1994). A distributed outstar network for spatial pattern learning. Neural Networks, 7, 159- 168. Technical Report CAS/CNS-TR-93-036, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/083_dOutstar_1994_.pdf
84. Carpenter, G.A. (1994). Distributed recognition codes and catastrophic forgetting. Proceedings of the World Congress on Neural Networks (WCNN-94), Hillsdale, NJ: Lawrence Erlbaum Associates, IV-133-142.
85. Carpenter, G.A. & Gjaja, M.N. (1994). Fuzzy ART choice functions. Proceedings of the World Congress on Neural Networks (WCNN-94), Hillsdale, NJ: Lawrence Erlbaum Associates, I-713-722. Technical Report CAS/CNS-TR-93-060, Boston, MA: Boston University.
86. Carpenter, G.A. & Ross, W.D. (1994). 3-D object recognition by the ART-EMAP evidence accumulation network. Proceedings of the World Congress on Neural Networks (WCNN-94), I-749-758. Technical Report CAS/CNS-TR-93-064, Boston, MA: Boston University.
87. Carpenter, G.A. & Grossberg, S. (1994). Fuzzy ARTMAP: A synthesis of neural networks and fuzzy logic for supervised categorization and nonstationary prediction. In R.R. Yager & L.A. Zadeh (Eds.), Fuzzy Sets, Neural Networks, and Soft Computing, New York: Van Nostrand Reinhold, 126-165.
88. Markuzon, N., Gaehde, S.A., Ash, A.S., Carpenter, G.A., & Moskowitz, M.A. (1994). Predicting risk of an adverse event in complex medical data sets using fuzzy ARTMAP network. Artificial Intelligence in Medicine: Interpreting Clinical Data. Technical Report Series, Menlo Park, CA: AAAI Press, 93-96.
89. Gaehde, S.A., Markuzon, N., Ash, A.S., Carpenter, G.A., & Moskowitz, M.A. (1994). Comparison of regression modeling and neural network modeling for predicting postoperative adverse events. Journal of General Internal Medicine, 9, 39.
90. Carpenter, G.A. (1994). Distributed recognition codes, catastrophic forgetting, and the rules of synaptic transmission. Proceedings: Intelligent Robots and Computer Vision XIII, Boston. SPIE Proceedings, 2353, Bellingham, WA: Society of Photo-Optical Instrumentation Engineers.
91. Carpenter, G.A. & Grossberg, S. (1994). Integrating symbolic and neural processing in a self-organizing architecture for pattern recognition and prediction. In V. Honavar & L. Uhr (Eds.), Artificial Intelligence and Neural Networks: Steps Toward Principled Integration, San Diego, CA: Academic Press, 387-421. Technical Report CAS/CNS TR-93-002, Boston, MA: Boston University.
92. Bradski, G., Carpenter, G.A., & Grossberg, S. (1994). STORE working memory networks for storage and recall of arbitrary temporal sequences. Biological Cybernetics, 71, 469-480. Technical Report CAS/CNS-TR- 92-028, Boston MA: Boston University.
93. Carpenter, G.A. & Grossberg, S. (1994). Self-organizing neural networks for supervised and unsupervised learning and prediction. In V. Cherkassky, J. H. Friedman, & H. Wechsler (Eds.), From Statistics to Neural Networks. Theory and Pattern Recognition Applications, ASI NATO Series F., 136. New York: Springer- Verlag, 319-348.
94. Carpenter, G.A. & Tan, A.-H. (1995). Rule extraction: From neural architecture to symbolic representation. Connection Science, 7, 3-27. Technical Report CAS/CNS-TR-94-005, Boston, MA: Boston University.
95. Carpenter, G.A. & Grossberg, S. (1995). Adaptive Resonance Theory. In M.A. Arbib (Ed.), The Handbook of Brain Theory and Neural Networks, Cambridge, MA: MIT Press, 79-82.
96. Carpenter, G.A. & Grossberg, S. (1995). A neural network architecture for autonomous learning, recognition, and prediction in a nonstationary world. In S.F. Zornetzer, J.L. Davis, C. Lau, & T. McKenna (Eds.), An Introduction to Neural and Electronic Networks, Second Edition, San Diego, CA: Academic Press, 465- 482. Technical Report CAS/CNS TR-93-049, Boston, MA: Boston University.
97. Carpenter, G.A. & Ross, W.D. (1995). ART-EMAP: A neural network architecture for object recognition by evidence accumulation. IEEE Transactions on Neural Networks, 6, 805-818. Technical Report CAS/CNS-TR- 93-035, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/097_ART-EMAP_1995_.pdf
98. Asfour, Y.R., Carpenter, G.A., & Grossberg, S. (1995). Landsat satellite image segmentation using the fuzzy ARTMAP neural network. Proceedings of the World Congress on Neural Networks (WCNN-95), I-150-156. Technical Report CAS/CNS TR-95-004, Boston, MA: Boston University.
99. Carpenter, G.A., Grossberg, S., & Reynolds, J.H. (1995). A fuzzy ARTMAP nonparametric probability estimator for nonstationary pattern recognition problems. IEEE Transactions on Neural Networks, 6, 1330- 1336. Technical Report CAS/CNS-93-047, Boston, MA: Boston University.
100. Carpenter, G.A. & Grossberg, S. (1996). Learning, categorization, rule formation, and prediction by fuzzy neural networks. In C.H. Chen (Ed.) Fuzzy Logic and Neural Network Handbook, New York: McGraw-Hill, 1.3- 1.45. Technical Report CAS/CNS TR-94-028, Boston, MA: Boston University.
101. Carpenter, G.A. (1996). Distributed activation, search, and learning by ART and ARTMAP neural networks. Proceedings of the International Conference on Neural Networks (ICNN'96): Plenary, Panel and Special Sessions, Piscataway, NJ: IEEE Press, 244-249. Technical Report CAS/CNS TR-96-006, Boston, MA: Boston University.
102. Carpenter, G.A., Gjaja, M.N., Gopal, S., & Woodcock, C.A. (1996). ART neural networks for remote sensing: Vegetation classification from Landsat TM and terrain data. International Geoscience and Remote Sensing Symposium, 1996, 1. Piscataway, NJ: IEEE Press, 529-531. Technical Report CAS/CNS TR-96- 008, Boston, MA: Boston University.
103. Carpenter, G.A. (1996). Distributed ART networks for learning, recognition, and prediction. Proceedings of the World Congress on Neural Networks (WCNN'96), Mahwah, NJ: Lawrence Erlbaum Associates, 333-344. Technical Report CAS/CNS TR-96-005, Boston, MA: Boston University.
104. Carpenter, G.A & Grossberg, S. (1996). Fuzzy ART. In B. Kosko, Fuzzy Engineering, New York: Prentice Hall, 467-497. Technical Report CAS/CNS-TR-93-059, Boston, MA: Boston University.
105. Carpenter, G.A. & Grossberg, S. (1996). Adaptive resonance theory: Self-organizing networks for stable learning, recognition, and prediction. In E. Fiesler & R. Beale (Eds.), The Handbook of Neural Computation, New York: Oxford University Press. Technical Report CAS/CNS-TR-95-017, Boston, MA: Boston University.
106. Carpenter, G.A. (1997). Spatial pattern learning, catastrophic forgetting, and optimal rules of synaptic transmission. In D.S. Levine & W.R. Elsberry (Eds.), Optimality in Biological and Artificial Networks? Mahwah, NJ: Lawrence Erlbaum Associates, 288-316. Technical Report CAS/CNS-TR-93-058, Boston, MA: Boston University.
107. Carpenter, G.A., Gjaja, M.N., Gopal, S., & Woodcock, C.E. (1997). ART neural networks for remote sensing: Vegetation classification from Landsat TM and terrain data. IEEE Transactions on Geoscience and Remote Sensing, 35, 308-325. Technical Report CAS/CNS-TR-95-026. Boston, MA: Boston University.
108. Carpenter, G.A. & Grossberg, S. (1997). Adaptive resonance theory. In J.D. Irwin (Ed.), The Industrial Electronics Handbook, Boca Raton, FL: CRC Press, 1286-1298.
109. Carpenter, G.A., Rubin, M.A., & Streilein, W.W. (1997). ARTMAP-FD: Familiarity discrimination applied to radar target recognition. Proceedings of the International Conference on Neural Networks (ICNN'97), 3, Piscataway, NJ: IEEE Press, 1459-1464. Technical Report CAS/CNS TR-96-032, Boston, MA: Boston University.
110. Grossberg, S., Carpenter, G., Schwartz, E., Mingolla, E., Bullock, D., Gaudiano, P., Andreou, A., Cauwenberghs, G., & Hubbard, A. (1997). Automated vision and sensing systems at Boston University. Proceedings of the DARPA Image Understanding Workshop, New Orleans, May.
111. Grossberg, S., Carpenter, G., Schwartz, E., Mingolla, E., Bullock, D., Gaudiano, P., Andreou, A., Cauwenberghs, G., & Hubbard, A. (1997). Principal investigator report: Automated vision and sensing systems at Boston University. Proceedings of the DARPA Image Understanding Workshop, New Orleans, May.
112. Carpenter, G.A. & Wilson, F.D.M. (1997). ARTMAP-DS: Pattern discrimination by discounting similarities. In W. Gerstner, A. Germond, M. Hasler, & J.-D. Nicoud (Eds.), Proceedingsof the International Conference on Artificial Neural Networks (ICANN'97), Berlin: Springer-Verlag, 607-612. Technical Report CAS/CNS-TR-97- 009, Boston, MA: Boston University.
113. Carpenter, G.A., Rubin, M.A., & Streilein, W.W. (1997). Threshold determination for ARTMAP-FD familiarity discrimination. In C.H. Dagli, M. Akay, O. Ersoy, B.R. Fernandez, & A. Smith (Eds.), Smart Engineering Systems: Neural Networks, Fuzzy Logic, Data Mining, and Evolutionary Programming, 7, (Proceedings of the Artificial Neural Networks in Engineering Conference - ANNIE'97), New York: ASME Press, 23-28. Technical Report CAS/CNS TR-97-006, Boston, MA: Boston University.
114. Carpenter, G.A., Gjaja, M.N., Gopal, S., Markuzon, N., & Woodcock, C.E. (1997). ART and ARTMAP neural networks for applications: Self-organizing learning, recognition, and prediction. In L.C. Jain (Ed.), Soft Computing Techniques in Knowledge-Based Intelligent Systems in Engineering, New York: Springer-Verlag, 279-317. Technical Report CAS/CNS TR-96-009, Boston, MA: Boston University.
115. Carpenter, G.A. (1997). Distributed
learning, recognition, and prediction by ART and ARTMAP neural
networks. Neural Networks, 10, 1473-1494. Technical
Report CAS/CNS TR-96-004, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/115_dART_NN_1997_.pdf
116. Carpenter, G.A., Grossberg, S., & Lesher, G.W. (1998). The what-and-where filter: A spatial mapping neural network for object recognition and image understanding. Computer Vision and Image Understanding, 69, 1-22. Technical Report CAS/CNS-TR-93-043, Boston, MA: Boston University.
117. Carpenter, G.A. & Markuzon, N. (1998). ARTMAP-IC and medical diagnosis: Instance counting and inconsistent cases. Neural Networks, 11, 323-336. Technical Report CAS/CNS-TR-96-017, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/117_ARTMAP-IC_1998_.pdf
118. Campos, M.M. & Carpenter, G.A. (1998). WSOM: Building adaptive wavelets with self-organizing maps. Proceedings of the International Joint Conference on Neural Networks (IJCNN'98), Piscataway, NJ: IEEE Press, 763-767. Technical Report CAS/CNS TR-97-007, Boston, MA: Boston University.
119. Carpenter, G.A. & Streilein, W.W. (1998). ARTMAP-FTR: A neural network for fusion target recognition, with application to sonar classification. AeroSense: Proceedings of SPIE's 12th Annual Symposium on Aerospace/Defense Sensing, Simulation, and Control. Orlando, April 13-17, 1998, Bellingham, WA: Society of Photo-Optical Instrumentation Engineers. Technical Report CAS/CNS TR-98-016, Boston, MA: Boston University.
120. Carpenter, G.A., Milenova, B., & Noeske, B. (1998). dARTMAP: A neural network for fast distributed supervised learning. Neural Networks, 11, 793-813. Technical Report CAS/CNS TR-97-026, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/120_dARTMAP_1998_.pdf
121. Martens, S., Gaudiano, P., & Carpenter, G.A. (1998). Mobile robot sensor fusion with fuzzy ARTMAP. Proceedings of the 1998 IEEE International Symposium on Computational Intelligence in Robotics and Automation (ISIC/CIRA/ISAS'98), Piscataway, NJ: IEEE Press, 307-312. Technical Report CAS/CNS TR- 98-011, Boston, MA: Boston University.
122. Streilein, W.W., Gaudiano, P., & Carpenter, G.A. (1998). A neural network for object recognition through sonar on a mobile robot. Proceedings of the 1998 IEEE International Symposium on Computational Intelligence in Robotics and Automation (ISIC/CIRA/ISAS'98), Piscataway, NJ: IEEE Press, 271-276. Technical Report CAS/CNS TR-98-015, Boston, MA: Boston University.
123. Campos, M.M. & Carpenter, G.A. (1998). Building adaptive basis functions with a continuous SOM. Proceedings of the Third International Conference on Computational Intelligence and Neuroscience (CI&N'98). Technical Report CAS/CNS TR-98-025, Boston, MA: Boston University.
124. Martens, S., Carpenter, G.A., & Gaudiano, P. (1998). Neural sensor fusion for spatial visualization on a mobile robot. Proceedings of the SPIE International Symposium on Intelligent Systems and Advanced Manufacturing (Boston, November, 1998). Technical Report CAS/CNS TR-98-028, Boston, MA: Boston University.
125. Carpenter, G.A., Gjaja, M.N., Gopal, S., & Woodcock, C.E., ART neural networks for remote sensing image analysis. Technical Report CAS/CNS TR-99-007, Boston, MA: Boston University.
126. Carpenter, G.A., & Milenova,
B.L. (1999). Distributed ARTMAP. Proceedings of the International
Joint Conference on Neural Networks (IJCNN'99), CD-ROM (IEEE
Catalog Number: 99CH36339C): #3022. Session 5.13. Technical Report
CAS/CNS TR-99-013, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/126_IJCNN_dARTMAP_1999_.pdf
127. Carpenter, G.A., Gopal, S., Macomber,
S., Martens, S., & Woodcock, C.E. (1999). A neural network
method for mixture estimation for vegetation mapping. Remote
Sensing of Environment, 70, 138-152. Technical Report
CAS/CNS TR-97-014, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/127_Mixtures_RSE_1999_.pdf
128. Carpenter, G.A., Gopal, S., Macomber, S., Martens, S., Woodcock, C.E., & Franklin, J. (1999). A neural network method for efficient vegetation mapping. Remote Sensing of Environment, 70, 326-338. Technical Report CAS/CNS TR-98-035, Boston, MA: Boston University.
129. Campos, M.M., & Carpenter, G.A. (2000). Building adaptive basis functions with a continuous self-organizing map. Neural Processing Letters, 11, 59-78. Technical Report CAS/CNS TR-99-005, Boston, MA: Boston University.
130. Carpenter, G.A. & Milenova,
B.L. (2000). ART neural networks for medical data analysis and
fast distributed learning. In Helge Malmgren, Magnus Borga, and
Lars Niklasson (Eds.), Artificial Neural Networks in Medicine
and Biology. Proceedings of the ANNIMAB-1 Conference, Göteborg,
Sweden, 13-16 May 2000, London: Springer-Verlag, 10-17. Springer
series Perspectives in Neural Computing. Technical Report
CAS/CNS TR-2000-002, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/130_ANNIMAB_2000_.pdf
131. Carpenter, G.A. (2000). ART neural
networks: Distributed coding and ARTMAP applications. In Peter
Sincák and Ján Vascák (Eds.), Quo Vadis
Computational Intelligence? New Trends and Approaches in Computational
Intelligence. In the series Studies in Fuzziness and Soft
Computing, New York: Physica-Verlag, 3-12. Technical Report
CAS/CNS TR-2000-005, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/131_ISCI_2000_.pdf
132. Kopco, N. & Carpenter, G.A. (2000). Graded signal functions for ARTMAP neural networks. In Peter Sincák, Ján Vascák, Vladimír Kvasnicka, & Radko Mesiar (Eds.), The State of the Art in Computational Intelligence. Proceedings of the European Symposium on Computational Intelligence (ISCI-2000), Kosice, Slovak Republic, August 30 September 1, 2000. In the series Advances in Soft Computing, New York: Physica-Verlag, 9-14. Technical Report CAS/CNS TR-2000-006, Boston, MA: Boston University.
133. Carpenter, G.A. (2000). Combining
distributed and localist computations in real-time neural networks.
A commentary on "Connectionist modelling in psychology: A
localist manifesto," by Mike Page. Behavioral and Brain
Sciences, 23, 473-474. Technical ReportCAS/CNS TR-99-028,
Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/133_BBS_Re_Page_2000_.pdf
134. Carpenter, G.A., (2000). Adaptive resonance: an emerging neural theory of cognition. Technical Report CAS/CNS TR-2000-010, Boston, MA: Boston University.
135. Carpenter, G.A. (2001). Neural
network models of learning and memory: leading questions and an
emerging framework. Trends in Cognitive Sciences, 5,
114-118. Technical Report CAS/CNS TR-2000-022, Boston, MA: Boston
University.
Download: http://cns.bu.edu/~gail/135_TiCS_2001_.pdf
136. Campos, M.M., & Carpenter,
G.A. (2001). S-TREE: Self-organizing trees for data clustering
and online vector quantization. Neural Networks, 14,
505-525. Technical Report CAS/CNS TR-2000-028, Boston, MA: Boston
University.
Download: http://cns.bu.edu/~gail/136_S-TREE_2001_.pdf
137. Shock, B.M., Carpenter, G.A., Gopal,
S., & Woodcock, C.E. (2001). ARTMAP neural network classification
of land use change. Proceedings of the World Congress on Computers
in Agriculture and Natural Resources, Iguaça Falls,
Brazil, September, 2001. Technical Report CAS/CNS TR-2001-009,
Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/137_Land_use_change_2001_.pdf
138. Carpenter, G.A., & Milenova, B.L. (2002). Redistribution of synaptic efficacy supports stable pattern learning in neural networks. Neural Computation, 14, 873-888. Technical Report CAS/CNS TR-99-019, Boston, MA: Boston University.
139. Waxman, A.M., Fay, D.A., Rhodes,
B.J., McKenna, T.S., Ivey, R.T., Bomberger, N.A., Bykoski, V.K.,
and Carpenter, G.A. (2002). Information fusion for image analysis:
Geospatial foundations for higher-level fusion. Proceedings
of the 5th International Conference on Information Fusion, Annapolis,
July.
Download: http://cns.bu.edu/~gail/139_Info_Fusion_2002_.pdf
140. Carpenter, G.A. & Grossberg, S. (2002). A self-organizing neural network for supervised learning, recognition, and prediction. In Thad A. Polk and Colleen M. Seifert (Eds.), Cognitive Modeling, Cambridge, MA: MIT Press, 288-314.
141. Carpenter, G.A. & Grossberg,
S. (2003). Adaptive Resonance Theory. In M.A. Arbib (Ed.), The
Handbook of Brain Theory and Neural Networks, Second Edition,
Cambridge, MA: MIT Press, 87-90. Technical Report CAS/CNS TR-98-029,
Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/141_HBTNN2e_ART_2003_.pdf
ARTICLES IN PRESS (5/2003)
Carpenter, G.A. (2003). Default ARTMAP.
Proceedings of the International Joint Conference on Neural
Networks (IJCNN'03), Portland, Oregon. Technical Report CAS/CNS
TR-2003-008, Boston, MA: Boston University.
Download: http://cns.bu.edu/~gail/Default_ARTMAP_2003_.pdf
(preprint, 1.8 MB)
Parsons, O., & Carpenter, G.A. (in
press). ARTMAP neural networks for information fusion and data
mining: Map production and target recognition methodologies. Neural
Networks. Technical Report CAS/CNS TR-2002-011, Boston, MA:
Boston University.
Download: http://cns.bu.edu/~gail/aARTMAP_map_TR-2002-011_3_.pdf
(preprint, 7MB) OR
http://cns.bu.edu/~gail/aARTMAP_map_NN_p-1_.pdf
+
http://cns.bu.edu/~gail/aARTMAP_map_NN_2003_.pdf (in press, 84
KB+ 744KB)
SUBMITTED FOR PUBLICATION (5/2003)
Carpenter, G.A., Gopal, S., Shock, B.M., & Woodcock, C.E., A neural network method for land use change classification, with application to the Nile River delta. Submitted to IEEE Transactions on Geoscience and Remote Sensing. Technical Report CAS/CNS TR-2001-010, Boston, MA: Boston University.
Kopco, N., & Carpenter, G.A., PointMap:
A real-time memory-based learning system with on-line and post-training
pruning. Technical Report CAS/CNS TR-2002-012, Boston, MA: Boston
University.
Download: http://cns.bu.edu/~gail/PointMap_TR-2002-012_.pdf
Olga Parsons (Ph.D., 2003)
Thesis: Neural Network Models for Spatial Data Mining, Map Production, and Cortical Direction Selectivity
Marcos M. Campos (Ph.D., 2002)
Thesis: Neural Networks and Adaptive Wavelets for Vector Quantization, Function Approximation, and Classification.
Boriana L. Milenova (Ph.D., 2000)
Thesis: Distributed Fast Learning in ART Neural Networks and an Application to Medical Database Analysis.
Siegfried Martens (Ph.D., 1999)
Thesis: Neural Networks for Satellite Remote Sensing and Robotic Sensor Interpretation
William W. Streilein (Ph.D., 1999)
Thesis: Neural Networks for Classification and Familiarity Discrimination, with Radar and Sonar Applications
Marin N. Gjaja (Ph.D., 1997)
Thesis: Neural Networks for Learning and Prediction with Applications to Remote Sensing and Speech Perception
Frank D.M. Wilson (Ph.D., 1996)
Thesis: Neural Networks for Noise-Tolerant Category Discrimination with Application to Continuous Speech Segmentation
Natalya Markuzon (Ph.D., 1996)
Thesis: Neural Networks for Supervised Learning and Prediction, with Applications to Character Recognition and Medical Database Analysis
Yousif R. Asfour (Ph.D., 1995)
Thesis: Fusion ARTMAP: Neural Networks for Multi-Sensor Fusion and Classification
Ah-Hwee Tan (Ph.D., 1994)
Thesis: Synthesizing Neural Network and Symbolic Knowledge Processing
William D. Ross (Ph.D., 1994)
Thesis: Neural Network Models of Attentive Visual Search and Object Recognition
John H. Reynolds (Ph.D., 1994)
Thesis: Neural Network Architectures for Learning, Prediction, and Probability Estimation
David B. Rosen (Ph.D., 1993)
Thesis: Fast Incremental Learning, Classification, and Prediction by Self-Organizing Neural Networks
John V. Jaskolski (Ph.D., 1992)
Thesis: Construction of Neural Network Expert Systems using Switching Theory
Adaptive Pattern Recognition and Categorization: Unsupervised Learning
(A) Analysis and development of ART 1, ART 2, and ART 3
This long-term research project has led to the development of the Adaptive Resonance Theory (ART) series of neural network modules that self-organize stable pattern recognition codes in real time in response to arbitrary sequences of analog or binary input patterns. New analysis and simulation studies of ART 1 [26] and ART 2 [31], as well as comparisons with biological and psychological data, have clarified the functions of these systems. ART 3 [40, 42], which includes a new type of analog search process, enhances the computational capabilities of ART systems so that they can be used with either fast learning or slow learning, and can robustly cope with asynchronous processing of arbitrary sequences of input patterns in real time. In ART 3 formal analogues of local neurotransmitter processes control global nonlinear feedback systems in the network. This work indicates solutions to computational bottlenecks which can occur when nonlinear feedback modules are embedded in neural network hierarchies. ART 1 is the central component of an airplane parts design and retrieval system at the Boeing Seattle plant (Caudell et al., 1991; Escobedo, Smith,& Caudell, 1993). There, the system is in current use in both design and manufacturing processes [39, 41, 43-46, 55, 58, 60, 61].
(B) Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system
A fuzzy ART model capable of rapid stable learning of recognition categories in response to arbitrary sequences of analog or binary input patterns has been developed. Fuzzy ART incorporates computations from fuzzy set theory into the ART 1 neural network, which learns to categorize only binary input patterns. The generalization to learning both analog and binary input patterns is achieved by replacing appearances of the intersection operator in ART 1 by the MIN operator of fuzzy set theory. The MIN operator reduces to the intersection operator in the binary case. Category proliferation is prevented by normalizing input vectors at a preprocessing stage. A normalization procedure called complement coding leads to a symmetric theory in which the MIN operator and the MAX operator of fuzzy set theory play complementary roles. Complement coding uses on- cells and off-cells to represent the input pattern, and preserves individual feature amplitudes while normalizing the total on-cell/off-cell vector. Learning is stable because all adaptive weights can only decrease in time. Decreasing weights correspond to increasing sizes of category "boxes." Smaller vigilance values lead to larger category boxes. Learning stops when the input space is covered by boxes. With fast learning and a finite input set of arbitrary size and composition, learning stabilizes after just one presentation of each input pattern. A fast- commit slow-recode option combines fast learning with a forgetting rule that buffers system memory against noise. Using this option, rare events can be rapidly learned, yet previously learned memories are not rapidly erased in response to statistically unreliable input fluctuations. Fuzzy ART forms part of a robot sensory- motor system under development at MIT Lincoln Laboratory (Bachelder, Waxman, & Seibert, 1993). [52, 56, 57, 85, 95, 104]
(C) ART 2-A: An adaptive resonance algorithm for rapid category learning and recognition
ART 2-A is an efficient algorithm that emulates the self-organizing pattern recognition and hypothesis testing properties of the ART 2 neural network architecture, but at a speed two to three orders of magnitude faster. Analysis and simulations show how the ART 2-A systems correspond to ART 2 dynamics at both the fast- learn limit and at intermediate learning rates. Intermediate learning rates permit fast commitment of category nodes but slow recoding, analogous to properties of word frequency effects, encoding specificity effects, and episodic memory. Better noise tolerance is hereby achieved without a loss of learning stability. The ART 2 and ART 2-A systems are contrasted with the leader algorithm. The speed of ART 2-A makes practical the use of ART 2 modules in large-scale neural computation. ART 2-A is the key component of the commercial software program Open Sesame (Charles River Analytics Inc.) that allows a Macintosh operating system to adapt to a user's work habits (Johnson, 1993). ART 2 and ART 2-A are also being used at MIT Lincoln Laboratory for face recognition (Seibert & Waxman, 1993); at Sandia National Laboratories, for target recognition (Moya, Koch, & Hostetler, 1993); and in Japan, for wave recognition in electrocardiograms (Suzuki, Yutaka, & Ono, 1993). [51, 53, patent 4]
(D) Normal and amnesic learning, recognition, and memory by a neural model of cortico-hippocampal interactions
The processes by which humans and other primates learn to recognize objects have been the subject of many models. Processes such as learning, categorization, attention, memory search, expectation, and novelty detection work together at different stages to realize object recognition. The structure and function of ART models are related to known neurological learning and memory processes, such as how inferotemporal cortex can recognize both specialized and abstract information, and how medial temporal amnesia may be caused by lesions in the hippocampal formation. The model also suggests how hippocampal and inferotemporal processing may be linked during recognition learning. [73]
Adaptive Pattern Recognition and Prediction: Supervised Learning
(A) ARTMAP: Supervised real-time learning and classification of nonstationary data by a self-organizing neural network
A neural network architecture, called ARTMAP, autonomously learns to classify arbitrarily many, arbitrarily ordered vectors into recognition categories based on predictive success. This supervised learning system is built up from a pair of ART modules that are capable of self-organizing stable recognition categories in response to arbitrary sequences of input patterns. During training trials, the ART module receives a stream of input patterns, and ART receives a stream of input patterns, where is the correct prediction given. These ART modules are linkedby an associative learning network and an internal controller that ensures autonomous system operation in real time. Tested on a benchmark machine learning database in both on-line and off-line simulations, the ARTMAP system learns orders of magnitude more quickly, efficiently, and accurately than alternative algorithms. The system conjointly maximizes generalization and minimizes predictive error by linking predictive success to category size on a trial-by-trial basis, using only local operations. This computation increases the vigilance parameter of ART by the minimal amount needed to correct a predictive error at ART. Rare but important events can be quickly and sharply distinguished even if they are similar to frequent events with different consequences. Because ARTMAP learning is self-stabilizing, it can continue learning one or more databases, without degrading its corpus of memories, until its full memory capacity is utilized. [47, 48, 50, 54, 59, patent 5]
(B) Fuzzy ARTMAP: A neural network architecture for incremental supervised learning of analog multidimensional maps
Fuzzy ARTMAP extends the capabilities of ARTMAP to carry out incremental supervised learning of recognition categories and multidimensional maps in response to arbitrary sequences of analog or binary input vectors. Improved prediction is achieved by training the system several times using different orderings of the input set. This voting strategy can also be used to assign confidence estimates to competing predictions given small, noisy, or incomplete training sets. Four classes of simulations illustrate fuzzy ARTMAP performance as compared to benchmark back propagation and genetic algorithm systems. These simulations include (i) finding points inside vs. outside a circle; (ii) learning to tell two spirals apart; (iii) incremental approximation of a piecewise continuous function; and (iv) a letter recognition database. Although recently introduced, fuzzy ARTMAP is already being applied to a variety of problems. Applications that have been described in the public domain include control of nuclear reactors (Keyvan, Durg, & Rabelo, 1993), medical database analysis (Ham & Han, 1993; Harvey, 1993), prediction of protein secondary structure (Mehta, Vij, & Rabelo, 1993), and land cover classification (Gopal, Sklarew, & Lambin, 1993). [63, 64, 68-70, 72, 80, 87, 91, 93, 96, 100, 105, 108]
(C) Fusion ARTMAP: A neural network architecture for multi-channel data fusion and classification
Fusion ARTMAP is a self-organizing neural network architecture for multi-channel, or multi-sensor, data fusion. Single-channel fusion ARTMAP is functionally equivalent to fuzzy ART during unsupervised learning and to fuzzy ARTMAP during supervised learning. The network has a symmetric organization such that each channel can be dynamically configured to serve either as a data input or a teaching input to the system. An ART module forms a compressed recognition code within each channel. These codes, in turn, become inputs to a single ART system that organizes the global recognition code. When a predictive error occurs, a process called parallel match tracking simultaneously raises vigilances in multiple ART modules until reset is triggered in one of them. Parallel match tracking hereby resets only that portion of the recognition code with the poorest match, or minimum predictive confidence. This internally controlled selective reset process is a type of credit assignment that creates a parsimoniously connected learned network. Fusion ARTMAP's multi-channel coding is illustrated by simulations of the benchmark Quadruped Mammal database. [74, 82]
(D) Comparative performance measures of fuzzy ARTMAP, learned vector quantization, and back propagation for handwritten character recognition
A simulation study comparesthe performance of fuzzy ARTMAP with that of learned vector quantization and back propagation on a handwritten character recognition task. Training with fuzzy ARTMAP to a fixed criterion uses many fewer epochs. Voting with fuzzy ARTMAP yields the highest recognition rates. One author of this paper (K. Iizuka) has returned to the Sharp Corporation where he is working on commercial applications of ART and ARTMAP systems. [66]
(E) Rule extraction by fuzzy ARTMAP Knowledge, in the form of fuzzy rules, can be derived from fuzzy ARTMAP at any stage of the learning process. However, the set of rules is often too large to be of practical value. An algorithm reduces the rule set to a manageable size. Rule extraction proceeds in two stages: pruning removes those recognition nodes whose confidence index falls below a selected threshold, and a quantization of continuous learned weights allows the final system state to be translated into a usable set of rules. Simulations on a medical prediction problem, the Pima Indian Diabetes (PID) database, illustrate the method. In the simulations, pruned networks about one-third the size of the original actually show improved performance. Quantization yields comprehensible rules with only slight degradation in test set predictive performance. [79, 94]
(F) Fuzzy ARTMAP, slow learning, and probability estimation
A nonparametric probability estimation procedure uses the fuzzy ARTMAP neural network. Because the procedure does not make a priori assumptions about underlying probability distributions, it yields accurate estimates on a wide variety of prediction tasks. Fuzzy ARTMAP is used to perform probability estimation in two different modes. In a "slow-learning" mode, input-output associations change slowly, with the strength of each association computing a conditional probability estimate. In "max-nodes" mode, a fixed number of categories are coded during an initial learning interval, and weights are then tuned by slow learning. Simulations illustrate system performance on tasks in which various numbers of clusters in the set of input vectors are mapped to a given class. [77, 99]
(G) Evaluation of speaker normalization methods for vowel recognition using fuzzy ARTMAP and K-NN
Fuzzy ARTMAP and K-Nearest Neighbor (K-NN) categorizers are used to evaluate intrinsic and extrinsic speaker normalization methods. Each classifier is trained on preprocessed, or normalized, vowel tokens from about 30 of the speakers of the Peterson-Barney database, then tested on data from the remaining speakers. Intrinsic normalization methods include one nonscaled, four psychophysical scales (bark, bark with end- correction, mel, ERB), and three log scales, each tested on four different combinations of the fundamental ( F0 ) and the formants ( F1, F2, F3 ). For each scale and frequency combination, four extrinsic speaker adaptation schemes are tested: centroid subtraction across all frequencies (CS), centroid subtraction for each frequency (CSi), linear scale (LS), and linear transformation (LT). A total of 32 intrinsic and 128 extrinsic methods are thus compared. Fuzzy ARTMAP and K-NN show similar trends, with K-NN performing somewhat better and fuzzy ARTMAP requiring about 1/10 as much memory. The optimal intrinsic normalization method is found to be bark scale, or bark with end-correction, using the differences between all frequencies (Diff All). The order of performance for the extrinsic methods is LT, CSi, LS, and CS, with fuzzy ARTMAP performing best using bark scale with Diff All, and K-NN choosing psychophysical measures for all except CSi. [76, 81]
(H) Neural networks for the analysis of satellite remote sensing images
A remote sensing test bed allows performance comparisonsbetween candidate neural network systems and traditional image processing and recognition techniques. In the course neural network of development, technology transfer to the remote sensing community feeds back to provide system design constraints. The first phase of this project used a data set collected in the Cleveland Forest, in southern California. ARTMAP networks were developed to identify vegetation classes based on input that includes six spectral bands (visible and IR) and terrain variables, such as aspect, slope, and elevation. Researchers working on the new NASA Earth Observing System (EOS) are now using this work to help design systems for image analysis, data compression, feature extraction, and temporal prediction for satellites scheduled to be launched in 1998 and beyond. A second phase of the remote sensing project began with planning the data collection for a study in the Plumas National Forest, in northern California. The new data set has been used to develop systems for accurate prediction of vegetation mixtures, in addition to single-class identification. A third phase of the project proves the practical utility of ARTMAP systems for large-scale mapping, by comparison with a standard, state-of-the-art method, mapping the Sierra National Forest. [98, 102, 107, 114, 125; Carpenter, Gopal, Macomber, Martens, Woodcock, & Franklin, in press]
(I) An ARTMAP system for predicting mixed categories applied to vegetation mapping
While most forest maps identify only the dominant vegetation class in delineated stands, individual stands are often better characterized by a mix of vegetation types. Many land management applications, including wildlife habitat studies, can benefit from knowledge of mixes. This project examines various algorithms that use data from the Landsat Thematic Mapper (TM) satellite to estimate mixtures of vegetation types within forest stands. Included in the study are maximum likelihood classification and linear mixture models as well as a new methodology based on the ARTMAP neural network. Two paradigms are considered: classification methods, which describe stand- level vegetation mixtures as mosaics of pixels, each identified with its primary vegetation class; and mixture methods, which treat samples as blends of vegetation, even at the pixel level. Comparative analysis of these mixture estimation methods, tested on data from the Plumas National Forest, yields the following conclusions: (1) accurate estimates of proportions of hardwood and conifer cover within stands can be obtained, particularly when brush is not present in the understory; (2) ARTMAP outperforms statistical methods and linear mixture models in both the classification and the mixture paradigms; (3) topographic correction fails to improve mapping accuracy; and (4) the new ARTMAP mixture system produces the most accurate overall results. The Plumas data set has been made available to other researchers for further development of new mapping methods and comparison with the quantitative studies presented here, which establish initial benchmark standards. [Carpenter, Gopal, Macomber, Martens, & Woodcock, in press]
(J) ARTMAP-IC and medical database analysis
Challenges of automated medical diagnosis include database size, missing values, noise, and inconsistent cases. The ARTMAP-IC [instance counting / inconsistent cases] network was developed in the context of these problems. During training, with winner-take-all code representation, ARTMAP-IC counts the number of cases assigned to each internal category. During testing, with a distributed category representation, output predictions are modulated by the instance counting layer to produce a probability prediction for each class. A new ARTMAP match tracking rule allows the network to encode inconsistent cases during training. Even when inconsistent cases are not a problem, the new match tracking rule increases code compression in the basic ARTMAP network by up to 50%. [88-89, 117].
(K) ARTMAP-FTR and sonar target classification
Special-purpose requirements of various application domains have led to a number of ARTMAP variants, including fuzzy ARTMAP, ART-EMAP, ARTMAP-IC, Gaussian ARTMAP, and distributed ARTMAP. This project develops a new ARTMAP variant, called ARTMAP-FTR (fusion target recognition), for the problem of multi- ping sonar target classification. The development data set, which lists sonar returns from underwater objects, was provided by the Naval Surface Warfare Center (NSWC) Coastal Systems Station (CSS), Dahlgren Division. The ARTMAP-FTR network has proven to be an effective tool for classifying objects from sonar returns. The system also provides a procedure for solving more general sensor fusion problems. [119]
(L) Sensor fusion and object recognition on a mobile robot
In another sonar application, ARTMAP systems have also been used for sensor fusion and object recognition on a mobile robot. [121, 122, 124]
Temporal Patterns, Working Memory, and 3-D Object Recognition
(A) Working memory networks for learning temporal order with application to 3-D object recognition
Working memory neural networks, called Sustained Temporal Order REcurrent (STORE) models, encode the invariant temporal order of sequential events in short-term memory (STM). An extension of the original STORE model permits sequences that include repeated items. Inputs to a STORE network may be presented with widely differing growth rates, amplitudes, durations, and interstimulus intervals without altering the stored invariant representation. The STORE temporal order code is designed to enable groupings of the events to be stably learned and remembered in real time, even as new events perturb the system. Such invariance and stability properties are needed in neural architectures which self-organize learned codes for variable-rate speech perception, sensory- motor planning, or 3-D visual object recognition. An enhanced STORE network permits accurate representation and recall of both repeated and non-repeated list items. [49, 62, 65, 92]
(B) ART-EMAP: Learning and prediction with spatial and temporal evidence accumulation
ART-EMAP is a neural architecture that uses spatial and temporal evidence accumulation to extend the capabilities of fuzzy ARTMAP. ART-EMAP combines supervised and unsupervised learning and a medium-term memory (MTM) process to accomplish stable pattern category recognition in a noisy input environment. The ART-EMAP system features (i) distributed pattern registration at a view category field; (ii) a decision criterion for mapping between view and object categories which can delay categorization of ambiguous objects and trigger an evidence accumulation process when faced with a low confidence prediction; (iii) a process that accumulates evidence at an MTM field; and (iv) an unsupervised learning algorithm to fine-tune performance after a limited initial period of supervised network training. The network, derived in four incremental stages, is applied to a benchmark 3-D object recognition problem to demonstrate temporal evidence accumulation. ART-EMAP systems, which feature winner-take-all codes during training but distributed codes during testing, provided a partial solution to the distributed coding problem addressed more generally by dART networks. [78, 86, 97]
Distributed Codes and Spatial Pattern Learning
(A) Distributed outstar learning and the rules of synaptic transmission
The distributed outstar, a generalization of the outstar neural network (Grossberg, 1968) for spatial pattern learning, has been developed. In the outstar, signals from a source node cause weights to learn and recall arbitrary patterns across a target field of nodes. The distributed outstar replaces the source node with a source field whose activity pattern may be arbitrarily distributed. Learning proceeds according to a principle of atrophy due to disuse, whereby a path weight decreases in joint proportion to the transmitted path signal and the degree of disuse of the target node. During learning, the total signal to a node converges toward that node's activity level. Weight changes are apportioned according to the distributed pattern of converging signals. Three synaptic transmission functions, a product rule, a capacity rule, and a threshold rule, are examined for this system. The three rules are computationally equivalent when source field activity is winner-take-all. When source field activity is distributed, catastrophic forgetting may occur. Only the threshold rule solves this problem. Analysis of spatial pattern learning by distributed codes thereby leads to the conjecture that the unit of long-term memory in such a system-and perhaps in many others-is an adaptive threshold, rather than the multiplicative path weight widely used in neural models. [75, 83, 84, 90, 106]
(B) Distributed ART architectures
Winner-take-all coding allows ART networks to maintain stable memories, but this type of code representation can cause problems such as category proliferation with fast learning and a noisy training set. A new class of ART models that permit arbitrarily distributed code representations has now been introduced. With winner-take-all coding, the unsupervised distributed ART model (dART) reduces to fuzzy ART and the supervised distributed ARTMAP model (dARTMAP) reduces to fuzzy ARTMAP. dART automatically apportions learned changes according to the degree of activation of each coding node, which permits fast as well as slow learning with compressed or distributed codes. Distributed ART models replace the traditional neural network path weight with a dynamic weight equal to the rectified difference between coding node activation and an adaptive threshold. Dynamic weights that project to coding nodes obey a distributed instar learning law and those that originate from coding nodes obey a distributed outstar learning law. Inputs activate distributed codes through phasicand tonic signal components with dual computational properties, and a parallel distributed match-reset-search process helps stabilize memory. A series of new studies to investigate computational properties of dART and dARTMAP systems are the basis of the new proposal. [101, 103, 115]
(C ) A distributed ARTMAP architecture
Distributed ARTMAP (dARTMAP) seeks to combine the computational advantages of MLP and ART systems in a realtime neural network for supervised learning. A dARTMAP implementation algorithm describes one class of these networks. This system incorporates elements of the unsupervised dART model as well as new features, including a contentaddressable memory (CAM) rule for improved contrast control at the coding field. A dARTMAP system reduces to fuzzy ARTMAP when coding is winnertakeall. Simulations show that dARTMAP retains fuzzy ARTMAP accuracy while significantly improving memory compression. [120]
(D) Redistribution of synaptic efficacy supports stable pattern learning in neural networks
Markram and Tsodyks (Nature, 1996), by showing that the elevated synaptic efficacy observed with single-pulse LTP measurements disappears with higher frequency test pulses, have critically challenged the conventional assumption that LTP reflects a general gain increase. Redistribution of synaptic efficacy is here seen as the local realization of a global design principle in a neural network for pattern coding. Namely, the network stabilizes learning by dynamically balancing a pattern-independent increase in synaptic strength against a pattern-specific increase in selectivity. This computation is implemented in the model by a monotonic long-term memory process which has a bidirectional effect on the postsynaptic potential via functionally complementary signal components. (Carpenter & Milenova, in press; subitted for publication)
Image Compression and Data Representation
This project has developed algorithms for image compression and efficient representation of noisy data. One target application of these systems is the rapid transmission of large-scale data sets. The new family of data compression systems employs multi-disciplinary computational structures and techniques, including neural networks, wavelets, and decision trees. These diverse methodologies are tied together by principles of self-organization and competitive learning, which have guided the construction of unified architectures for complex function representation tasks. New systems whose performance has been successfully compared to that of state-of-the-art algorithms include: WSOM (Wavelet Self-Organizing Map), CSOM (Continuous Self-Organizing Map), and S-TREE (Self-Organizing Tree). All three methods use on-line training. Although each learning rate decreases with time, it is possible to reactivate learning to process nonstationary data and to perform adaptive vector quantization (VQ) tasks.
(A) Wavelet Self-Organizing Map (WSOM)
WSOM has been implemented in unsupervised and supervised algorithms. The latter can be used, in the context of vector quantization, to develop compression systems where labels of the training set can be used to enforce greater compression in less critical areas of the data. For example, in medical image compression, regions that are not relevant can be compressed further yielding poorer reconstruction without risk. The code is written in MATLAB. WSOM is able to create, in a concise way, multidimensional wavelet bases that preserve many of the desirable properties of the discrete wavelet transform. Once these wavelet bases/frames have been created for a given type of data, they can be used for compressing large multidimensional data sets. In general, the creation of multidimensional wavelet bases/frames have relied upon radial or ridge-like wavelets and they do not allow the user to take advantage of the standard kernels used in the low-dimensional cases. WSOM provides a fast way of creating high-dimensional bases and uses the same machinery of the low-dimensional wavelet approaches. This is a promising approach for producing better compression schemes for high-dimensional data sets. [118]
(B) Continuous Self-Organizing Map (CSOM)
CSOM is implemented as a supervised
system, but the algorithm could also be used in an unsupervised
setting. It has been used to create models for function approximation
and classification tasks. The code is written in MATLAB. CSOM
provides a continuous version of the Self-Organizing Map (SOM).
Besides applications in regression, where it has been shown to
perform better than SOM and as well as radial basis function (RBF)
on test examples, CSOM can be used in vector quantization. In
this case the position in the grid provides a low-dimensional
representation. The system then uses these coordinates for reconstructing
the data. This approach promise to yield smoother data reconstruction
than current methods. [123; Campos & Carpenter, in press]
Invariant Visual Preprocessing
(A) A What-and-Where neural network for invariant image preprocessing
The What-and-Where filter is a feedforward neural network for invariant image preprocessing that represents the position, orientation, and size of an image figure (where it is) in a multiplexed spatial map. This map is used to generate an invariant representation of the figure that is insensitive to position, orientation, and size for purposes of pattern recognition (what it is). A multiscale array of oriented filters, followed by competition between orientations and scales is used to define the Where filter. [67, 116]
RESEARCH SUMMARY REFERENCES
Bachelder, I.A., Waxman, A.M., and Seibert, M. (1993). A neural system for mobile robot visual place learning and recognition. Proceedings of the World Congress on Neural Networks (WCNN-93), I -512-517. Hillsdale, NJ: Lawrence Erlbaum Associates.
Caudell, T., Smith, S.D.G., Johnson, C., Wunsch, D., and Escobedo, R. (1991). An industrial application of neural networks to reusable design. Adaptive Neural Systems, Technical Report BCS-CS-ACS-91-001, Seattle, WA: The Boeing Company, 185-190.
Escobedo, R., Smith, S.D.G., and Caudell, T.P. (1993). The ART of design retrieval. Adaptive Neural Systems, Technical Report BCS-CS-ACS-93-008, Seattle, WA: The Boeing Company, 149-160.
Gopal, S., Sklarew, D.M., and Lambin, E. (1993). Fuzzy-neural network classification of land cover change in the Sahel. Proceedings of the DOSES/EUROSAT Workshop on New Tools for Spatial Analysis. Lisbon, Portugal, November 18-20, 1993.
Ham, F.M. and Han, S.W. (1993). Quantitative study of the QRS complex using fuzzy ARTMAP and the MIT/BIH arrhythmia database. Proceedings of the World Congress on Neural Networks (WCNN-93), I -207-211. Hillsdale, NJ: Lawrence Erlbaum Associates.
Harvey, R.M. (1993). Nursing diagnosis by computers: An application of neural networks. Nursing Diagnosis, 4, 26-34.
Johnson, C. (1993). Agent learns user's behavior. Electrical Engineering Times, June 28, 43, 46.
Keyvan, S., Durg, A., and Rabelo, L.C. (1993). Application of artificial neural networks for development of diagnostic monitoring system in nuclear plants. American Nuclear Society Conference Proceedings, April 18- 21, 1993.
Markram, H., and Tsodyks, M. (1996). Redistribution of synaptic efficacy between neocortical pyramidal neurons. Nature, 382, 807-810.
Mehta, B.V., Vij, L., and Rabelo, L.C. (1993). Prediction of secondary structures of proteins using fuzzy ARTMAP. Proceedings of the World Congress on Neural Networks (WCNN-93), I -228-232. Hillsdale, NJ: Lawrence Erlbaum Associates.
Moya, M.M., Koch, M.W., and Hostetler, L.D. (1993). One-class classifier networks for target recognition applications. Proceedings of the World Congress on Neural Networks (WCNN-93), III -797-801. Hillsdale, NJ: Lawrence Erlbaum Associates.
Seibert, M. and Waxman, A.M. (1993). An approach to face recognition using saliency maps and caricatures. Proceedings of the World Congress on Neural Networks (WCNN-93), III -661-664. Hillsdale, NJ: Lawrence Erlbaum Associates.
Suzuki, Y., Abe, Y., and Ono, K. (1993) Self-organizing QRS wave recognition system in ECG using ART 2. Proceedings of the World Congress on Neural Networks (WCNN-93), IV -39-42. Hillsdale, NJ: Lawrence Erlbaum Associates.
Boston-area seminar in mathematical biology (1976-1978)
AMS Special Session (Philadelphia, 1980): Unsolved problems in biological and chemical modeling
Center for Adaptive Systems and Department of Cognitive and Neural Systems, Boston University: Co-organizer of ongoing in-house and public seminars and lectures (1980-present)
Symposium honoring Sonia Kovalevsky, Cambridge, MA, October 1985 (organizing committee)
SIAM/AMS Committee on Mathematics in the Life Sciences (1983-1986): Organizer, 1986 Annual AAAS Symposium: Modeling Circadian Rhythms, Philadelphia
2-day meeting on Pattern Recognition in Natural and Artificial Neural Systems, Society for Mathematical Biology, June, 1986, Boston (Co-organizer)
2-day NSF workshop: Neural Networks and Neuromorphic Systems, October, 1986, Woburn, MA (Co- organizer)
IEEE International Conference on Neural Networks, June, 1987, San Diego (program committee)
IEEE Conference on Neural Information Processing Systems-Natural and Synthetic, November, 1987, Denver, Colorado (program committee)
International Neural Network Society (INNS), Annual Meeting, September, 1988, Boston (Organization Chairman; program committee)
International Joint Conference on Neural Networks (IJCNN), January, 1990, Washington, DC (program committee)
Wang Institute of Boston University: Course (Neural Networks: From Foundations to Applications) and Research Conference (Neural Networks for Automatic Target Recognition), May 6-13, 1990, Tyngsboro, MA (Co- Chairman)
Wang Institute of Boston University: Course (Neural Networks: From Foundations to Applications) and Research Conference (Neural Networks for Vision and Image Processing), May 5-12, 1991, Tyngsboro, MA (Co- Chairman)
International Conference on Artificial Neural Net (ICANN-91), June, 1991, Helsinki, Finland (program committee)
FUZZ-IEEE-92: IEEE International Conference on Fuzzy Systems. March 8-12, 1992, San Diego, CA (program committee)
Wang Institute of Boston University Course (Neural Networks: I Introduction and Foundations. II Research and Applications) and Research Conference (Neural Networks for Recognition, Learning, and Control), May 10- 16, 1992, Tyngsboro, MA (Co-Chairman)
International Joint Conference on Neural Networks (IJCNN), June, 1992, Baltimore, Maryland (program committee)
Second International Conference on Fuzzy Logic and Neural Networks, July, 1992, Iizuka, Japan (program committee)
Russian Neural Network Society (RNNS)/IEEE Joint Conference on Neural Networks. October 7-10, 1992, Rostov-on-Don, Russia (program committee)
NSF Workshop on Neuroengineering. October, 1992, College Park, Maryland (program committee)
World Congress on Neural Networks (WCNN)/INNS Annual Meeting. July 11-15, 1993, Portland, Oregon (program committee)
World Congress on Neural Networks (WCNN)/INNS Annual Meeting. June 4-10, 1994, San Diego, CA (program committee)
Turkish Symposium on Artificial Intelligence and Neural Networks (TAINN'94) Ankara, Turkey (program committee)
Third International Conference on Fuzzy Logic, Neural Nets and Soft Computing, August, 1994, Iizuka, Japan (program committee)
World Congress on Neural Networks (WCNN)/INNS Annual Meeting. July 17-21, 1995, Washington, DC (program committee)
International Conference on Pattern Recognition, August 25-30, 1996, Vienna, Austria (program committee)
World Congress on Neural Networks (WCNN) / INNS Annual Meeting. September 15-20, 1996. San Diego, CA (program committee)
International Conference on Neural Information Processing (ICONIP'96) October 22-25, 1996, Hong Kong (program committee)
International Conference on Artificial Neural Networks (ICANN'97) October, 1997, Lausanne, Switzerland (program committee)
International Joint Congress on Neural Networks (IJCNN'98) May, 1998, Anchorage, Alaska (program committee)
International Conference on Artificial Neural Networks (ICANN'98) September, 1998, Skövde, Sweden (program committee)
International Joint Congress on Neural Networks (IJCNN'99) July, 1999, Washington, DC (program committee)
International Conference on Artificial Neural Networks (ICANN'99) September, 1999, Edinburgh, UK (program committee)
International Symposium on Computational Intelligence (ISCI - 2000) August - September, 2000, Kosice, Slovakia (honorary chairman)
CONFERENCE ORGANIZATION (2001-)
International Joint Congress on Neural Networks (IJCNN'01) July, 2001, Washington, DC (program committee)
Euro-International Symposium on Computational Intelligence (ISCI - 2002) June, 2002, Kosice, Slovakia (program committee)
International Joint Congress on Neural Networks (IJCNN'02) May, 2002, Honolulu, Hawaii (program committee)
International Joint Congress on Neural
Networks (IJCNN'03) July, 2003, Portland, Oregon (program committee)