Keyword Index 
a-Si:H, 763 
abstraction, 385 
action hierarchies, 385 
activation functions, 279 
active example selection, 593 
active learning, 231,287, 385,443,593,705, 
1069 
active memory, 141 
active vision, 451,893 
actor/critic algorithms, 401 
adaptive control, 1031 
adaptive critic, 401 
adaptive interface, 843 
adaptive optimal control, 361 
advantage updating, 353 
affine transformation, 713, 1101 
AIFNN, 1101 
alternating minimization, 569 
amorphous silicon memory, 763 
amplification, 3 
analog VLSI, 731,755,787, 795, 803,817 
analog-digital VLSI, 811 
analog VLSI synapses, 763 
annealing schedule, 965 
approximate dot product, 747 
arm movements, 1117 
ART 1,755 
articulatory loop, 51 
associative memory, 35, 77, 149, 513, 925 
associative net, 513 
attractor networks, 11, 35, 149 
attractors, 173 
auditory localization chip, 787 
auditory models, 827 
auditory system, 125 
autoassociative MLP, 991 
autoassociative networks, 941 
autoencoder networks, 165 
autonomous vehicle navigation, 657 
averaging, 419 
awareness, 11 
axon circuit, 739 
backpropagation, 191, 1031, 1055 
barn owl, 125 
basis functions, 157 
Bayesian inference, 255 
Bellman residual, 353 
bias, 295 
binary features, 475 
biological modeling, 61 
biophysical modeling, 3 
blind separation, 467 
block codes, 795 
Boltzmann learning, 435, 617 
boosting, 875 
bootstrap, 1055 
breast cancer, 1063 
cable theory, 69 
capacity information efficiency, 513 
CART, 1063 
cascade-correlation, 537, 601 
catastrophic forgetting, 419 
catastrophic interference, 19 
centroid, 999 
chaotic time series, 311 
character recognition, 747 
chemosensory receptors, 69 
chess, 1069 
circular normal density, 641 
City Block multilayer perceptron, 747 
classification, 1109 
cleanup process, 11 
clustering, 361,459, 585, 625, 681,713,779, 
999 
CMOS, 709, 811 
CMOS vector quantizer, 779 
coarse-to-fine search, 981 
cochlear model, 827 
1138 Keyword Index 
cocktail party, 467 
cognitive models, 141 
cognitive task, 11 
combined estimators, 295,419 
combining sensory modalities, 787 
committee, 231,419 
compartmental models, 69 
competition, 199 
competitive clustering, 561 
competitive learning, 85, 475, 497 
computational complexity, 183 
computational models, 85 
conditional densities, 641 
conditional probability, 1047 
confidence, 1055 
confidence intervals, 489 
connectionist expert systems, 505 
consciousness, 11 
consolidation, 19 
constrained rank Gaussian mixtures, 681 
constructive learning, 537 
context, 981 
continuous speech recognition, 859 
continuous time, 353 
control, 19, 361, 1007, 1031 
control gain, 1031 
convergence, 585,649 
convergence of backpropagation, 335 
convergence of stochastic gradient 
descent, 335 
convolution network, 901 
cooperation, 199 
cooperative vector quantization, 617 
coordinate transformations, 1125 
coronary artery bypass, 1055 
correlation based learning, 93 
correlation network, 909 
correlogram, 827 
cortical columns, 117 
cortical plasticity, 109 
cosine units, 537 
cost functions, 529 
credit assignment through time, 427, 553 
cross connections, 601 
cross-validation, 231,489 
cursive handwriting recognition, 1109 
data sampling, 593 
data selection, 459 
data transformations, 223 
decision processes, 345 
decision theory, 1077 
decoding with dynamics, 795 
decorrelation, 475, 925 
deformable model, 965 
delay line, 739 
delusions, 149 
dementia, 35 
density estimation, 419, 641, 1047 
desynchrony, 199 
diagnosis, 1077 
differential games, 353 
direct multi-step prediction, 721 
direction selectivity, 3 
discrete-time, 883 
distance measures, 713 
distributed representations, 505, 577 
dopamane, 101,141 
dynamic decay adjustment, 521 
dynamac modelling, 311 
dynamic programming, 345, 369, 393, 401 
dynamac writing transformation, 1093 
dynamical decorrelation, 925 
dynamical systems, 481, 1031 
early stopping, 263,489 
effective complexity, 263 
effective machine size, 263 
electronic transformation, 69 
electrontonic space, 69 
electrotonus, 69 
EM algorithm, 419, 427, 435, 545, 553, 
585,617, 681,713,835,957, 1077, 1085 
ensemble, 231 
entropy, 287, 467,475 
error feedback, 1031 
ERS- 1 satellite, 641 
estimation, 295 
evidence, 255 
example selection, 593 
expectation-maximization, 419,427,435, 545, 
553,585, 617, 681,713,835, 957, 1077, 
1085 
experiment design, 593 
explanation-based neural networks, 1069 
extra outputs, 657 
face recognition, 747 
facial expressions, 909 
factoffal coding, 467 
factorial learning, 561,617 
feature maps, 577 
Keyword Index 113 9 
feedback, 925 
feedback connections, 3 
financial markets, 411,529 
finite size effects, 217 
finite state machine, 697 
fixed point attractors, 319 
flat minima, 529 
floating gate synapses, 817 
focal lesions, 35 
forecasting, 721 
foreign exchange, 411 
forward model, 43 
frame of reference, 157 
function approximation, 231,361 
fusion, 1007 
gain control, 925 
games, 1069 
Gamma operator, 883 
Gauss-Newton, 673 
Gaussian mixture models, 1077 
generalization, 207, 215, 223,263,287, 303, 
327, 361,489, 529, 649 , 657, 1125 
generalization dynamics, 263 
gesture recognition, 901 
gesture-to-speech device, 843 
Gibbs sampling, 617 
global motion perception, 917 
Glove-Talk II, 843 
gradient descent, 585,803 
grammar, 27 
grammar inference, 427 
graph matching, 713 
graph searching, 665 
ground truth, 1085 
grow-support algorithm, 369 
growing structures, 497 
H-infinity training, 191 
hallucinations, 149 
hand tracking, 901 
handwriting decomposition, 1023 
handwriting preprocessing, 1093 
handwriting synthesis, 1023 
handwritten character recognition, 965, 
1093, 1101 
handwritten digit recognition, 999 
head direction, 173 
heart arrythmia, 731 
heart attack, 1055 
Hebbian learning, 93, 319 
Hebbian synapses, 69 
Helmholtz machine, 1015 
hidden Markov models, 427, 435, 553, 617, 
851,859, 867, 875 
hidden state, 377 
hierarchical mixture of experts, 835 
hierarchical representation, 941 
hierarchies, 385 
high dimensional spaces, 747 
hints, 411 
hippocampal neurons, 69 
hippocampus, 173 
hippocampus model, 77 
HMM, 427, 435, 553,617, 851,859, 867, 875 
honey bee, 61 
horizontal connections, 109 
Hough transform, 965 
human-computer interaction, 901, 1039 
hybrid digital-analogue network, 1007 
hyperparameters, 255 
hysteresis, 3 
mage 
mage 
mage 
image 
mage 
mage 
analysis, 1085 
compression, 941 
interpolation, 973 
model, 965 
processing, 451,901 
pyramids, 941 
mage recognition, 657 
mage search, 981 
mperfectly learnable problems, 287 
mplantable systems, 731 
incomplete data, 545 
incomplete patterns, 689 
information gain, 287, 475 
nformation theory, 467, 925 
inhibition, 61,141 
nstance-based learning, 377 
instantiation parameters, 965 
interacardiac electrogram classifier, 731 
interference, 1117 
nterior point method, 569 
nternal model, 43 
mvariances, 991, 1007 
nvariant distance, 999 
invariant recognition, 223 
inverse dynamics, 19 
inverse problems, 101 
iterated prediction, 311 
iterative projective scaling, 957 
1140 Keyword Index 
joint probability distribution, 1077 
JPMAX, 933 
k-blocking, 279 
k-means, 585, 681 
k-nearest neighbor, 377 
Kalman filter, 43 
Kerr-type materials, 771 
keyboard input, 1039 
knowledge incorporation, 991 
knowledge representations, 601 
Kohonen map, 497, 893 
Korean speech recognition, 811 
Lagrangian, 771 
lateral connections, 117 
lateral geniculate, 133 
lateral inhibition, 109 
learning curves, 239, 303, 327 
learning dynamics, 207 
learning from hints, 411 
learning rate, 303 
least squares, 295 
lesions, 35 
LGN organization, 133 
Lie groups, 223 
limit cycles, 481 
limiting performance, 239 
linear perceptron, 207 
linear programming, 505 
Linsker model, 319 
lipreading, 851,973 
LISSOM, 109 
local error bars, 835 
local linear models, 973, 1015 
local models, 697 
local PCA, 973 
locally weighted regression, 705 
LOESS, 705 
log-polar sensors, 893 
logistic regression, 1055, 1063 
long-term dependencies, 427 
low loss redundancy reduction, 247 
low power neural net, 731 
LVQ, 681 
Lyapunov exponent, 311 
macaque monkey, 93 
magnetic equilibrium, 1007 
manifold learning, 973 
Markov decision problems, 345, 393 
Markov decision process, 361 
Markov distributions, 279 
Markov Monte Carlo, 443 
master equations, 271 
maximum entropy, 459, 545 
maximum expected utility, 1077 
maximum likelihood, 489, 545, 633 
mean field theory, 957 
medical risk, 1055 
medicine, 1063 
memory, 51 
memory-based leaming, 377 
minimax, 353 
minimax estimation, 191 
minimum description length, 165, 1015 
missing data, 459, 545,689, 1077 
mixture models, 419, 459, 585, 641,681,689, 
705, 713, 1015, 1077 
mixture of experts, 19, 419, 427, 633, 843, 
859, 973, 1125 
model selection, 443 
model-fitting, 933 
modular architecture, 901 
modularity, 11 
moments, 271 
Monte-Carlo, 345 
morphogenesis, 133 
motion, 933 
motion processing, 165 
motor control, 43, 1017, 1125 
motor learning, 19, 1023, 1117 
motor maps, 893 
movement primitives, 1023 
MST, 165 
multi-resolution, 981 
multidimensional scaling, 459 
multiple attractors, 319 
multiple experts, 1085 
multiple outputs, 721 
multitask learning, 657 
n-of-m expressions, 609 
natural language, 27 
navigation, 173 
nearest neighbor, 377, 999 
nearest sequence memory, 377 
network analysis, 601 
network interpretation, 505, 609 
neural coding, 165 
neural control, 1031 
Keyword Index 1141 
neural development, 93 
neural gas, 497, 625 
neural predictor, 1047 
neuroanatomy, 117 
neuro-chess, 1069 
neuromodulation, 141 
neuromorphic rendering, 69 
neuronal modeling, 3 
neuroscience, 125 
Newton optimization, 585 
NMDA synapse, 101 
noise, 649 
noisy data, 689 
noisy neurons, 279 
non-equilibrium systems, 271 
non-Gaussian limiting behavior, 335 
non-Markov environments, 377 
nonlinear control, 1031 
nonlinear decorrelation, 247 
nonlinear dynamics, 319 
nonlinear interpolation, 973 
nonlinear system identification, 883 
object recognition, 901,933 
observer model, 43 
OCR, 965,991, 1015, 1093, 1101, 1109 
ocular dominance, 93 
ocular dominance wavelength, 109 
ocular-motor learning, 1125 
olfaction, 61 
on-line learning, 303, 335 
optic flow, 165 
optical back-propagation, 771 
optical character recognition, 665, 991, 
1015, 1093, 1101, 1109 
optical imaging, 93 
optical implementation, 771 
optimal brain surgeon, 673 
optimal stopping, 263 
orientation, 173 
orientation adaptation, 925 
orientation columns, 925 
orientation selectivity, 93, 117 
oscillator, 199 
outcome, 1063 
overfitting, 489, 529 
PAC learning, 215, 279 
pairwise classifier, 1109 
pancakes, 1015 
parallel implementations of MLP, 747 
parameter estimation, 545 
parameter fitting, 61 
parietal cortex, 157 
Parzen windows, 689 
patchy projections, 117 
pathological attractors, 149 
pattem classification, 665 
pattem model, 93 
pattem recognition, 999 
PCA-pyramids, 941 
perceptron, 279, 303 
perceptual learning, 917 
perceptual organization, 165 
performance prediction, 239 
periodic variables, 641 
persistent state, 11 
perturbative learning, 803 
phase space learning, 481 
phonological store, 51 
piecewise fits, 633 
plasma, 1007 
plasticity, 173 
plasticity site, 125 
plasticity-mediated competition, 475 
Poisson clumping heuristic, 327 
pose, 957 
posterior probabilities, 545, 1109 
prediction, 489, 1063 
prediction error, 443 
predictive coding, 1047 
prefrontal cortex, 141 
primary visual system, 319 
principal component analysis, 601,941 
prior knowledge, 411,419, 553,713 
probabilistic classifier, 1109 
probabilistic interpretation, 489 
probabilistic outputs, 1109 
product units, 537 
protein sequences, 459 
pruning, 529, 673 
psychophysical models, 917 
psychophysics, 19 
pulse-based computation, 739 
Q-leaming, 353, 361,385, 393,401 
qualitative analysis, 319 
quantization, 941 
quatemions, 957 
queries, 443,593,705 
query learning, 231,287 
queueing systems, 393 
114 2 Keyword Index 
radial basis functions, 497, 521,569, 1125 
rank deficient mixtures, 681 
rat, 173 
RBF, 521 
real-time recurrent learning, 673 
real-time control, 1007 
receiver operating characteristic, 875, 1085 
receptive field formation, 319 
receptor-transducer model, 61 
recurrent networks, 141, 31 I, 427, 553, 649, 
697 
redundancy reduction, 247 
refractory effects, 271 
regression, 231,489 
regularization, 223,649 
rehearsal, 51 
reinforcement learning, 101, 125,345, 361, 
369, 377, 385, 393, 401, 1069 
relaxation, 199 
remote sensing, 641, 1085 
representation, 361 
residual gradient algorithms, 353 
robot arm kinematics, 505 
robustness, 191 
ROC, 875, 1085 
rotation invariance, 665 
rule extraction, 505, 601,609 
rule verification, 505 
saccadic eye movements, 893 
saliency map, 451 
sample complexity, 215, 327 
sample size requirements, 327 
scene segmentation, 199 
schizophrenia, 141,149 
second messenger, 61 
second-order methods, 529, 585 
second-order optimization, 673 
segmentation, 165, 901 
selective attention, 451 
selective gating, 199 
self-organization, 27, 77, 109, 319,925 
semi-Markov decision problems, 393 
sense of direction, 173 
sensorimotor coordination, 157 
sensorimotor integration, 43, 1125 
sequence analysis, 933 
sequence learning, 427, 553 
sequence processing, 577 
serial order, 51 
shared hidden units, 721 
shift operator, 883 
short-term memory, 51,141 
signal processing, 883 
single transistor synapses, 817 
singular value decomposition, 999 
skills, 385 
smooth value iteration, 369 
soft clustering, 957 
soft assign, 957 
softmax, 51 
spatial filters, 893 
spatial representations, 157 
speaker adaptation, 867 
speaker normalization, 867 
speaker transformation, 875 
spectrogram inversion, 827 
speech coding, 835 
speech processing, 467 
speech reading, 851 
speech recognition, 577, 859, 867, 875 
speech recognition with mixture of 
experts, 859 
speech reconstruction, 827 
speech representation, 827 
spikes, 739 
spiking neurons, 183 
state aggregation, 361 
state estimation, 43 
state identification, 377 
statistical mechanics, 271,287, 545 
statistical physics, 255 
statistical properties of high-D spaces, 747 
stochastic approximation, 303, 335, 803 
stochastic gradient ascent, 101 
stochastic learning, 649 
stochastic neurodynamics, 271 
stochastic perceptrons, 279 
strabismus, 109 
structure, 385 
subjective probability, 1085 
subspace models, 999 
sun spots, 673 
superior colliculus, 893 
supervised leaming, 255, 657, 721 
survival, 1063 
symbolic rules, 609 
symmetries, 223 
symmetry hint, 411 
synaptic interactions, 69 
synaptic perturbation, 101 
synchrony, 199 
Keyword Index 1143 
system identification, 697, 883 
talker transformation, 875 
talking machine, 843 
tangent distance, 665, 991,999, 1015, 1101 
TD leaming, 721 
templates, 957 
temporal coherence, 933 
temporal difference, 401 
temporal difference learning, 1069 
temporal dynamics, 263 
temporal processing, 85,883 
temporal sequences, 51 
text compression, 1047 
text prediction, 1039, 1047 
thalamus, 173 
thermodynamic limit, 207 
three-dimensional object recognition, 949 
three-state neurons, 271 
time series, 489 
time series prediction, 435,673,721,835 
time-delay neural network, 697, 1093 
timing precision, 183 
TNM staging system, 1063 
tokamak, 1007 
topological map, 497, 577 
trajectory learning, 481 
transfer, 19 
transformation, 991 
transformation invariance, 665 
translation invariance, 665 
two-dimensional views, 949 
two-port analysis, 69 
two-spirals problem, 601 
typing, 1039 
unrealizable rule, 255,287 
unsupervised 3-D object recognition, 949 
unsupervised learning, 109, 165, 247,459, 
467, 475, 497, 561,577, 585, 617, 
625, 681,713,933,949 
unsupervised motor learning, 1023 
URAN, 811 
validity interval analysis, 505 
value iteration, 369 
van der Pol oscillator, 481 
variance, 295 
variance estimation, 705 
VC dimension, 215, 327 
VC bounds, 327 
vector quantization, 625, 779, 835 
vectorial code, 157 
velocity tuning, 3 
virtual examples, 411 
vision, 451,933,973 
visual attention, 451 
visual cortex, 3, 93, 157, 165 
visuomotor learning, 1125 
VLSI, 731,811 
VLSI algorithms, 747 
VLSI dynamical decoder, 795 
VLSI synapses, 817 
vocabulary acquisition, 51 
VQ chip, 779 
weight decay, 529, 649 
weight perturbation, 803 
weight templates, 609 
well-formed state, 11 
wind vectors, 641 
winner-take-all, 51,513,681 
word spotting, 875 
working memory, 1117 
zip codes, 999 
