KEYWORD INDEX 
A-current, 566 
a priori knowledge, 1001 
abstraction levels, 1178 
action potential, 927 
activation overlap, 1176 
active learning, 391,679 
adaptation, 927 
adaptive 
control, 647, 1077 
filtering, 351,559 
grids, 663 
momentum, 477 
pattern recognition, 833 
routing, 671 
address block location, 745,785 
address reading, 785 
AIC, 293 
algebraic energy functions, 1184 
co-transformation, 271 
amplification, 519 
analog, 311 
retrieval, 1109 
VLSI, 582, 850, 858, 874, 927 
VLSI chips, 769 
analogical similarity, 1109 
analysis techniques, 1117 
analytic continuation, 335 
analyzing wavelet, 423 
annealing, 896 
approximation power, 319 
architecture, 335 
area MT, 969 
arithmetic comparison, 1117 
artificial intelligence, 1143 
assemblies, 463 
associative memory, 919, 1125 
asymptotic convergence, 391 
asynchronous, 493 
attention, 1167 
attractor neural networks, 485,493 
attractors, 75,527 
audition, 1069, 1163 
auditory processing, 606 
auditory scene analysis, 1069 
auto-associative nets, 152 
autoencoder, 3,271 
autoencoder networks, 11 
autonomous navigation, 655 
averaging, 1188 
Bach, J. S., 1163 
backpropagation, 232, 351, 1161 
backpropagation convergence, 383 
backpropagation-correction term, 200 
backpropagation simulation, 888 
barn owl, 614 
Bayes approach, 977 
Bayesian, 1001 
inference, 208 
learning, 200 
model, 590, 606 
networks, 285 
updating, 485 
bee, 527 
best choice problem, 801 
bias variance, 367 
BIC, 293 
binary diamond, 1143 
binary weights, 359 
binding problem, 993, 1109 
biological modeling, 574, 904, 1077, 1167 
Boltzmann, 896 
learning algorithm, 896 
machine, 27, 83 
Boolean networks, 1143 
boosting, 1188 
Brownian motion, 83 
bumptree, 240 
C++, 843 
capacity, 375 
capacity control, 208, 303 
1195 
1196 Keyword Index 
cascade-correlation with cross-validation, 793 
catastrophic forgetting, 1176 
catastrophic interference, 1176 
cellular neural nets (CNN), 888 
center-surround lateral connectivity, 629 
chaos, 647 
character recognition, 216, 911,937 
chemistry, 208 
choice of activation function, 319 
circuit complexity, 359 
classification, 120, 136, 343, 1143 
clustering, 19, 27, 96, 184, 501,590, 809 
coding, 463 
cognitive maps, 1101 
cognitive modeling, 1085, 1167 
color, 136 
committee machine, 399 
committee of networks, 208 
committees, 1188 
communication delay, 493 
comparing function, 801 
competitive learning, 112 
competitive neural networks, 104 
complex 
analysis, 335 
cell, 953 
distance metrics, 168 
complexity, 391,439 
componential code, 1085 
componential representation, 27 
compositional hierarchy, 285 
compression, 144 
computational efficiency, 1178 
computer vision, 1182 
conditional independence, 285 
confidence intervals, 415 
connection machine, 904 
connectionist hardware, 1178 
connectionist modeling, 1178 
constrained supervised learning, 1043 
constraint 
boxes, 882 
satisfaction, 896 
constructive leaming, 88, 279, 1178 
content and co-content functions, 882 
context, 1180 
context-dependence, 75, 1180 
context modelling, 1051 
continuous representation, 19 
continuous speech recognition, 1059 
continuous-time dynamics, 858 
contractive system, 493 
contrast adaptation, 769 
contribution analysis, 1117 
control, 647, 719, 1169 
convergence proofs, 703 
convergence rate, 477 
convergent networks, 1184 
convolutional neural network, 745, 937 
coordinate descent, 96 
correspondence problem, 961,985 
cortical map, 543 
cost function, 200, 1019 
counting function, 375 
coupled dynamics, 447 
covering problems, 1184 
critical phenomena, 439 
cross-connections, 1117 
cross-correlations, 629 
cross-validation, 59, 391 
cursive handwriting recognition, 833 
data clustering, 104 
decision tree, 240, 911, 1035 
deficient data, 128 
density estimation, 120, 961 
detection, 1019 
deterministic annealing, 96, 985 
deterministic Boltzmann machine, 896 
development, 543, 1001 
dichotomy, 375 
differential equation, 423 
diffusion networks, 83 
digital circuits, 911 
digital signal processor (DSP), 888 
dimension independent bounds, 319 
dimension reduction, 152 
diophantine equations, 431 
discontinuities, 977 
discrete gradient, 232 
discrete representation, 19 
discretization, 501 
discriminant learning, 1035 
discriminative models, 825 
Keyword Index 1197 
discriminative training, 1019 
distance measure, 96, 152 
distortion measure, 152 
distributed implementation, 493 
distributed representations, 3, 11, 1109 
divide and conquer, 1180 
DNA pattern recognition, 761 
document processing, 745 
domain assumptions, 1169 
dopamine, 559 
DRAM, 843 
drug activity prediction, 216 
dual neural network, 801 
dynamic 
networks, 850, 1011 
programming, 590, 639, 663,703 
reposing, 216 
systems, 719 
time warping, 945 
dynamics, 447 
echo suppression, 1069 
effective complexity, 303 
effective number of parameters, 35 
eigenspace decomposition, 263 
eigenvector, 144 
elastic matching, 769 
electromyography (EMG), 1043 
EM, 120, 128, 937 
EM algorithm, 192 
encoder problem, 144 
encoders, 1101 
ensemble dynamics, 463 
ensembles, 1188 
entraining, 1163 
entropy, 19, 271 
elYor 
bars, 208 
correcting codes, 777 
functions, 647 
propagation, 455 
evidence procedure, 200 
evolutionary algorithms, 88 
exemplar selection, 391 
expectation-maximization, 96 
exploration, 160, 679, 1169 
eye movement, 945 
face recognition, 769 
facial feature tracking, 753 
factorial codes, 3 
factorial representation, 27 
factorization, 1180 
fan-in restrictions, 359 
fault detection, 825 
fault-prone software modules, 793 
fault tolerance, 455 
feature 
detector, 745 
extraction, 136, 785 
manifold problem, 216 
maps, 817 
selection, 200 
figure/ground discrimination, 993 
figure of merit, 1019 
finite state machine, 19,359, 501 
firing rates, 463 
Fisher information matrix, 293 
fitness landscapes, 51 
foraging, 598 
forgetting, 1176 
forward and inverse relaxation model, 1 043 
forward dynamics model, 647 
forward modelling, 679 
free energy, 3 
frequency identification, 271 
frequency normalization, 953 
FSM extraction, 501 
function learning, 311 
gain, 874 
gain control, 551 
game playing, 817 
gamma memory, 1011, 1051 
gamma model, 1011 
gap-junction, 559 
Gaussian 
classifier, 793 
mixtures, 19, 120, 128, 825 
networks, 850 
synapses, 485 
gaze tracking, 753 
GDS, 1093 
gene modeling/genome modeling, 761 
gene parsing, 761 
1198 Keyword Index 
generalization, 35,240, 255,263,271, 311, 
327, 343, 1176 
dynamics, 303 
error, 367, 375,399 
unbiased estimation of, 391 
generalized cross-validation, 415, 1059 
generalized Hebbian algorithm, 144 
genetic algorithms, 51 
gesture recognition, 945,961 
global training, 937 
global trajectory optimization, 663 
Go, 817 
Hoo optimality, 351 
hand gesture recognition, 945 
handwriting model, 727 
handwriting recognition, 777 
handwritten character recognition, 727 
hardware implementation, 843 
hardware learning, 232 
heating, 574 
Hebbian learning, 407 
Hessian, 263 
hidden Markov model, 75, 83,719,761,825, 
937, 1051, 1059 
hidden unit noise, 1101 
hierarchical 
filtering, 168 
learning, 655 
structure, 1109 
high-performance simulation, 888 
high-risk modules, 793 
higher order statistics, 136 
Hilbert's tenth problem, 431 
hill climbing, 51 
history-dependent dynamics, 485 
Hodgkin-Huxley model, 566 
Hoeffding's bound, 59 
Hopfield network, 485, 1125 
human-computer interaction, 753 
human genes/human genome, 761 
human memory, 1085 
hybrid methods, 1188 
hyperbolic, 455 
hypercube, 904 
ICEG (intra cardiac electrogram), 874 
illusory contours, 993 
image 
compression, 104 
processing, 911,945, 1182 
segmentation, 745,993 
understanding, 1143 
IMAX, 809 
implementation, 858, 911 
in loop training, 874 
Incomplete data, 120, 128 
incremental learning, 255, 1178 
independent opinion pooling, 1027 
indirect adaptive methods, 695 
influence function, 192 
information criterion, 293 
information theory, 271, 551 
nput modality, 753 
insect, 527 
Integrate-and-fire models, 629 
integrate-and-fire neural network, 535 
ntegrated mean squared error, 391 
integrated segmentation and recognition, 1027 
Interactive simulator, 888 
interference, 1176 
intermediate-level vision, 993 
internal representation, 271,614, 1101 
Interneuron, 535 
nvafiance, 817 
Invariant object recognition, 769 
nverse dynamics model, 1043, 1077 
k-d trees, 590, 711 
k-nearest neighbors, 168 
k-satisfaction, 439 
Karhunen-Lo6ve expansion, 136 
kernel hidden units, 271 
kernel regression methods, 1165 
knot placement, 247 
Kohonen network, 843 
Lagrange multipliers, 96 
Laguerre memory, 1011 
landmark learning, 1101 
language models, 176 
lateral inhibition, 535 
layout analysis, 785 
Keyword Index 1199 
learning, 1182 
algorithms, 75,311,351,911, 1161 
complexity of, 1161 
control, 160, 663, 711 
curves, 327 
dynamics, 407 
from examples, 311 
structural, 88 
supervised, 1182 
theory, 176 
vector quantization (LVQ), 112 
line segment matching, 985 
linear networks, 144 
lipreading, 43, 1027 
LMS, 351,477, 1161 
loading problem, 431 
local 
field potential, 629 
kNN, 184 
learning, 184, 1165 
linear models, 152, 160 
minimum, 423 
models, 1180 
principal components, 43, 152 
time, 493 
trajectory optimization, 663 
localization, 1069 
locally recurrent network, 1051 
locally weighed regression, 160 
low-risk modules, 793 
macaque, 543 
machine vision, 753 
Mackey-Glass, 850 
manifolds, 43 
MAP, 200 
MAP estimation, 19 
Markov 
decision problems, 687, 695 
decision processes, 703 
models, 176 
random fields, 977 
match networks, 285 
maximum entropy estimation, 104 
maximum likelihood estimation, 3, 120, 423, 
679 
mean field annealing, 1184 
mean field theory, 882, 896, 977, 985 
memory-based learning, 59 
memory-based methods, 1165 
memory efficiency, 375 
memory retrieval, 1109 
minimal TP optimal control, 639 
minimization principle, 727 
minimizing disagreement, 112 
minimum description length (MDL), 3, 11, 
293,833 
missing data, 128 
missing features, 120, 961 
mixing function, 27 
mixture 
distribution, 192 
models, 120 
of experts, 43,719, 1180, 1188 
model-based recognition, 285 
model-based vision, 285 
model matching, 96, 985 
model merging, 1051 
model of neural system, 606 
model selection, 59, 192, 303,327, 343 
modeling, 519 
modular architecture, 719, 817 
momentum, 477 
monotone system, 493 
Monte Carlo algorithms, 687 
motion, 977 
motion parallax, 969 
motion planning, 655 
motion priorities, 614 
motoneuron, 535 
motor control, 144, 614, 1043 
motor learning, 1077 
MT filter, 969 
multi-agent learning, 671 
multi-layer classifier, 793 
multi-layer perception, 248 
multidimensional scaling, 104 
multiple causes, 27 
multiplierless, 232 
muscle, 535 
MUSIC, 888 
music, 1163 
music cognition, 1085 
musk odor prediction, 216 
mutual information, 809, 911, 1001 
1200 Keyword Index 
N-best paradigm, 1059 
natural images, 551 
nearest neighbor, 184, 843, 1165 
neocortex, 519 
NET32K processor, 785 
network 
complexity, 303,367 
dynamics, 75, 493 
simplification, 927 
size, 303,359 
neural computation, 904 
neural net simulator, 888 
neural networks, complexity of, 1161 
neural tree network, 1035 
neurocontrol, 647 
neurodynamical system, 455 
neuromodulation, 559 
neuromodulator, 598 
neuron, 527 
neuron MOS transistor, 919 
neuron simulator, 927 
NEXUS simulator, 953 
noise, 455 
noise sensitivity signature (NSS), 343 
noisy data, 128 
non-linear dynamics, 407 
nonmonotone 
convergence, 383 
dynamics, 485 
optimization, 383 
nonparametric procedure, 343 
nonparametric regression, 160, 247 
novelty detection, 825 
NP-complete, 1161 
object 
localization, 985 
recognition, 745,961, 1182 
objective function, 647 
observability, 335,455 
Observers' Paradox, 501 
occlusions, 977 
Ockham factors, 208 
ocular dominance, 543 
oculomotor system, 582 
olfaction, 527 
on-chip learning, 896 
on-line 
backpropagation convergence, 383 
character recognition, 937 
learning, 184, 477, 825 
training, 566 
word recognition, 777 
I/f noise, 629 
one-shot learning, 1143 
optic tectum, 606 
optical flow, 977 
optical imaging, 543 
optimal 
brain damage, 263 
brain surgeon, 263 
control, 639, 655,663,703 
convergence, 477 
experiment design, 679 
signalling, 485 
size neural networks, 343 
optimality, 1184 
optimization, 51,407, 1184, 1188 
orientation selectivity, 543 
orthogonalization, 144, 614 
oscillations, 463,629,866 
overfitting, 343,590 
overlap of representations, 1176 
overtraining, 263 
owl, 606 
PAC-learning, 311 
packet routing, 671 
parallel 
backpropagation, 383 
implementation, 843 
machines, 1178 
supercomputer, 888 
parameter estimation, 566 
pattern formation, 629 
pattern recognition, 945 
pen-based computing, 737 
penalized log likelihood, 415 
penalty terms, 1093 
perception classifier, 793 
perceptual vividness, 993 
performance prediction, 327 
perturbation, 455 
perturbed gradient, 383 
Keyword Index 1201 
phase-locking, 866 
phase transition, 439 
phases of learning, 303 
phoneme timing estimation, 727 
phonetic modelling, 1051 
piecewise-linear classifier, 112 
pitch, 1085 
point matching, 985 
poles, 335 
population codes, 11 
practical TPDP, 639 
precedence effect, 1069 
preceptive field surround, 969 
prediction, 343,598, 1163 
prediction suffix trees, 176 
predictive Hebbian learning, 598 
pretraining, 1176 
principal components, 27, 43,407 
principal components analysis (PCA), 35, 136, 
152, 1117 
principal components pruning, 35 
prior knowledge, 825 
prioritized sweeping, 695 
probability estimation, 961 
probalistic automata, 833 
programming environments, 1178 
projection pursuit, 1059 
protein secondary structure, 809 
pruning, 35,200, 208, 263, 1035 
pruning algorithm, 293 
psychophysics, 953 
pulsed neural networks, 927 
pyramidal cells, 519 
Q-learning, 639, 671,703 
Q-routing, 67 l 
quantization, 19, 232 
querying, 679 
RAAM model, 1125 
radial basis function, 240, 255, 319, 423,647, 
843,850, 961, 1165 
random k-CNF, 439 
rate code, 463 
RC networks, 882 
real-time dynamic programming (RTDP), 687, 
695 
real-time learning, 858 
real-time vision, 753 
receptive field, 1077 
recognition-based segmentation, 745,777 
recurrent inhibition, 535 
recurrent network, 75, 88, 279, 359, 431,501, 
566, 719, 858, 1051, 1085, 1180 
reduced-order control, 614 
regression, 35 
regularization, 35, 1059 
reinforcement learning, 639, 655, 663,671, 
687, 695,703, 711, 817, 1169 
remote sensing, 850, 1143 
Renshaw cell, 535 
replica method, 399 
replicas, 439 
representation, 1085 
reproducing kernel Hilbert space, 415 
rescheduling, 801 
resource allocating network (RAN), 1165 
response pattern, 527 
retina model, 559 
retinal processing, 769 
retrieval of stored pattern, 375 
reverberation suppression, 1069 
risk 
statistical, 391 
unbiased, 415 
robot 
control, 655 
learning, 160 
navigation, 711 
robotics, 679, 1077, 1169 
robust learning, 655 
robust regression, 192 
robustness, 351 
routing, 671 
rule-based networks, 1143 
rule generation, 1093 
saccade, 582 
scale invariance, 551 
scheduling, 801 
second-order methods, 263 
segmental neural network, 1059 
segmentation, 745 
selective attention, 1180 
1202 Keyword Index 
self-learning neural network, 919 
self-organization, 247, 255, 1001 
self-organizing feature maps, 104 
sensitivity to initial conditions, 501 
sensory integration, 606, 1027 
sequence learning, 75 
sequence recognition, 75 
shadowing, 455 
shape from texture, 953 
short-term memory, 1011 
shortest paths, 671 
Siamese neural network, 737 
sigmoidal functions of high order, 319 
sigmoidal input/output function, 485 
signal processing, 590 
signature verification, 737 
silicon retina, 769 
SIMD, 843 
simulated annealing, 1184 
smulation, 904 
sngle cells, 519 
singular value decomposition, 614 
singular values, 144 
singularities, 335 
smoothing spline, 415 
Sobolev spaces, 319 
soft classification, 415 
soft-hardware logic, 919 
softmax, 882 
solvable model, 423 
sound localization, 574, 1069 
sound separation, 1069 
space displacement neural network, 937 
sparse networks, 904 
spatial cognition, 1101 
spatial frequency, 953 
speaker recognition, 1035 
spectroscopy, 208 
speech, 1019 
articulator, 1043 
processing, 1035 
reading, 1027 
recognition, 1019, 1027, 1051 
synthesis, 1043 
spike, 463 
spike sorting, 590 
spiking neurons, 629 
spin glass, 439 
splice junction recognition, 1093 
spline analysis of variance, 415 
stacking, 1188 
static complexity metrics, 793 
statistical 
grammar, 1059 
mechanics, 104, 399,407, 882 
physics, 977 
stochastic 
approximation, 703,858 
gradient descent, 477 
learning, 471 
models, 833 
networks, 83 
stomatogastric ganglion, 566 
structure-form-motion, 969 
superior colliculus, 582 
supersmoothing, 160 
surface interpolation, 969 
surface learning, 43 
surface perception, 993 
switched capacitor, 874 
symbol manipulation, 1125 
synapse circuit, 919 
synchrony, 535 
tangent distance, 168, 216, 1165 
tangent prop, 216 
target tracking, 866 
teacher forcing, 566 
template matching, 919 
temporal 
difference, 687, 703, 817 
pattern, 463, 1011 
sequence, 1085 
texture compression, 953 
three-layered perception, 423 
threshold logic, 359 
threshold logic units, 375 
time-delay neural network, 737, 1027 
time series, 825, 1163 
time series prediction, 850, 1093 
topographic maps, 11 
topographic relations, 1101 
trainable gain, 874 
training data, 1182 
