Keyword Index 
2-step learning, 1071 
a priori, 957 
action potential, 166 
active learning, 528 
Adaboost, 647 
adaptation, 712 
adaptive basis functions, 402 
adaptive filtering, 712 
adaptive grid, 1036 
adaptive metric, 245 
adaptive problem-solving, 444 
admission control and routing, 922 
advantage learning, 1022 
aerial images, 689 
affine invariance, 843 
affine transfort-nation, 829 
agent, 1043 
aircraft recovery, 1022 
algorithm, 371 
analog, 726 
analog VLSI, 880 
analogy, 682 
annealed learning, 301 
anomaly detection, 943 
antisymmetric, 329 
arbitrary functions, 619 
architectural limits, 619 
area V1,236 
artifact identification, 229, 894 
asset allocation, 936 
associative indexing, 675 
associative memory, 52, 654, 675 
asymptotic theory, 294 
asynchronous circuits, 705 
asynchronous processes, 1064 
attention, 3, 80 
attractor dimension, 315 
attractor networks, 66, 329 
audiovisual fusion, 742 
auditory, 124 
auditory computation, 180 
auditory perception, 3 
auditory streaming, 3 
autoassociator, 472 
autoassociator network, 647 
autonomous navigation, 859 
autonomous robots, 822 
AVSR, 742 
axon guidance, 152, 159 
Bach, 957 
basic block, 929 
basis functions, 24 
Bayes, 689,829 
Bayesian fusion, 742 
Bayesian inference, 395, 402, 964 
Bayesian methods, 577, 787 
Bayesian networks, 479, 521, I001 
BCM, 423 
behavior, 1043 
belief networks, 479, 486, 584 
Bellman-equation, 1036 
bias learning, 245 
bidirectional retrieval, 675 
binocular disparity, 208 
binocular rivalry, 187 
biomedical application, 964 
biophysical modeling, 208 
bipartite graph matching, 780 
biped robot, 1071 
blind separation, 563,696, 756, 894 
blind source separation, 229 
Boltzmann machines, 280, 598 
bond rating, 661 
boolean logic, 252 
boolean neuron, 252 
boosting, 647 
bottom-up, 605 
brain damage, 66 
catastrophic fusion, 742 
central clustering, 514 
cerebellum, 13 8 
channel coding, 479 
character recognition, 647, 843 
character segmentation, 94 
chemotropism, 152, 159 
classical conditioning, 117 
classification, 500, 542, 591,887, 964, 978 
classification trees, 542 
classifier, 507 
cleaning, 992 
clustering, 409, 465 
CMAC, 1071 
CMOS, 726 
CMOS imager, 873 
cognitive modeling, 66 
columnar structure, 52 
combination of neural networks, 472 
combining models, 668 
combining nodels, 507 
combining predictions, 451,591 
committee machine, 378 
competition, 350 
1084 Keyword Index 
competitive models, 742 
complex cell, 208, 236, 801 
complexity, 371, 385 
complexity theory, 252 
computational complexity, 194 
computational limits, 619 
computer security, 943 
computer vision, 528, 726 
conditional independence, 563 
conditionally positive functions, 343 
connected speech, 166 
constrained optimization, 1022 
constraint dictionary, 689 
constructive learning, 131, 535 
context, 950 
context modeling, 742 
context monitoring, 742 
context-based recognition, 94 
context-free languages, 87 
continuous attractor, 654 
continuous state-space, 1036 
control and planning, 1015 
convergence rate, 1064 
convergent algorithm, 1029 
convolution, 908 
cooperation, 350 
coordinate leaxaing, 145 
correspondence analysis, 591 
cortex, 124 
counting, 87 
coupling, 507 
covariance matrix, 640 
cross-validation, 668 
cyclopean view, 808 
data analysis, 521 
data clustering, 528 
data compression, I0 
data dependent SRM, 336 
decision trees, 259, 336, 514 
dendritic processing, 208 
density esfrnation, 556, 570, 668, 815 
density expansions, 273 
deterministic annealing, 430, 514 
Diabetes Mellitus, 971 
diabolo network, 647 
differential entropy, 273 
diffusion, 159 
digit recognition, 416 
digital implementation, 705 
dimensionality reduction, 10, 626, 633,749 
discounted MDPs, 1064 
discretization, 1036 
discrimination, 507 
discfiminative training, 763 
disparity, 208 
disparity map, 808 
diverse density, 570 
divisive normalization, 173 
domestic chick, 31 
dopamine, 117 
drifting dynamics, 735 
drug activity prediction, 570 
dual-route models, 59 
dynamic programming, 1015, 1050, 1057 
dynamical systems, 87, 315 
early vision, 187 
edge detection, 873 
EEG, 735, 894 
efficiency and robustness, 385 
electroencephalographic, 894 
EM, 486, 780, 850, 971 
EM algorithm, 430, 500, 584, 626, 668 
ensemble learning, 395 
ensemble methods, 266 
ensembles, 591 
Eph receptors, 152 
equivariant property, 696 
error bars, 493 
error coding, 542 
eror estimates, 1036 
error model, 971 
eror-correction, 479 
errors-in-variables, 992 
event related potentials, 901 
evidence framework, 964 
example driven retrieval, 866 
expectation, 859 
exponential size, 252 
exponentiated gradient, 287 
extended self-similarity, 836 
extra inputs, 437 
extrapolation, 682 
face detection, 472 
face recognition, 17, 843 
facilitation, 194 
factor analysis, 45, 486, 626, 633,749 
factorial decision processes, 1057 
fading, 756 
familiarity, 73 
fault detection, 859 
feature discovery, 682 
feature selectivity, 124, 794 
feature transformation, 763 
filter model, 236 
financial analysis, 936, 992 
finite training-set, 357 
finite-difference method, 1029 
finite-state machine, 1015 
first order motion, 801 
Fisher information, 173,385, 696 
fly vision, 880 
focal-plane processor, 873 
fractals, 836 
fraud detection, 943 
Keyword Index 1085 
FSA, 619 
function approximation, 535 
functional optimization, 322 
Gabor filter, 726 
gamma distribution, 985 
Gaussian, 350 
Gaussian processes, 493 
generalization, 259, 294 
generalization analysis, 336 
generalization error, 357 
generalive models, 131,605, 654 
Gibbs sampling, 493 
global analysis, I0 
goldfish, 138 
gradient descent, 287 
gradients, 152, 159 
graph matching, 689 
graphical models, 416, 521,563 
greedy optimization, 131,929 
Green's functions, 343 
growth cone, 159 
habimation, 138 
Hamilton-Jacobi-Bellman equation, 1029 
Hamiltonian, 329 
hand tracking, 859 
handwriting recognition, 647 
handwritten digits, 605 
hardware image processing, 873 
harmony, 957 
hearing, 103 
Hebbian learning, 31,549, 675 
Helmholtz machine, 131 
heteroscedastic noise, 493 
hidden Markov models, 598, 749, 763 
hidden variables, 500 
hierarchical mixture model, 780 
hierarchical segmentation, 514 
hierarchy, 689, 1043, 1050 
high dimensional models, 45 
Hilbert Schmidt kernels, 343 
hippocampus, 73,145 
human reading, 94 
human vision, 829 
hybrid monte carlo, 964 
hybrid system, 763, 1071 
hypercolunm, 173 
ICA, 894 
ideal observer, 829 
dentification and classification, 901 
image analysis, 836 
mage classification, 570 
mage compression, 430 
image database, 866 
image de-noising, 773 
mage recognition, 773 
image reconstruction, 794 
image reU'ieval, 866 
imprinting, 31 
incremental learning, 612 
Independent Component Analysis, 229, 273, 556, 
563, 815, 894 
induction, 45 
inferior colliculus, 180 
inferotemporal cortex, 215 
infomax, 756 
information capacity, 675 
information retrieval, 451,528, 675 ' 
information theory, 103,201 
ranate behavior, 31 
insect vision, 822, 880 
nstmction scheduling, 929 
InSUlin metabolism, 971 
ntegrated service networks, 922 
interference, 73 
internal representation, 378 
interpolation, 682 
totersegmental coordination, 719 
intrusion detection, 943 
mvariances, 215,640 
Japanese OCR, 245 
Kalman filter, 80, 458, 971 
kernels, 640 
knowledge representation, 45 
Kronecker product, 696 
Kullbaek-Leibler divergence, 266, 395, 402, 794 
kurtosis, 423 
LANDSAT images, 514 
latent semantic analysis, 45 
latent variables, 222 
lateral connections, 486 
LDA, 763 
leanting, 45, 280, 500, 957, 992 
learning rule, 385 
learning rules, 194 
learning theory, 259 
learning with query, 612 
limit cycle, 329 
linear autoencoders, 626 
linear functions, 500 
linear regression, 364 
local regression, 633 
local trajectory optimization, 1008 
logistic regression, 287 
low-density parity-check codes, 479 
low-resolution OCR, 94 
LSA, 45 
Lyapunov, 329 
machine learning, 444, 661 
1086 Keyword Index 
Magnetoencephalography, 229 
manifold learning, 682 
Markov decision problems, 922 
Markov decision processes, 1050, 1057 
Markov networks, 521 
Markov process, 1043 
Markov Random Fields, 486 
Markovian sequences, 465 
master equations, 301 
matched filtering, 222, 901 
matching loss, 287 
matrix inverse, 385 
maximal margin hyperplane, 336 
maximum entropy, 273 
maximum likelihood, 696, 756 
Maximum Spanning Tree algorithm, 584 
MCMC, 577 
MDP, 1043 
mean field theory, 280, 416 
medicine, 950 
memory, 73 
memory relxieval, 52 
method of scoring, 696 
nunimum cross-entropy, 514 
mismatch negativity, 3 
rmssing data, 626, 971 
rmxture distribution, 416 
rmxture models, 222, 577, 584, 850 
rmxtures of experts, 668 
MLP, 763 
MMI, 763 
mobile communications, 756 
model comparison, 815 
model selection algorithms, 437, 444 
model-based learning, 1008 
modeling, 173,915 
modular networks, 1071 
modularity, 17 
modulation transfer function, 180 
molecular confirmation, 570 
monitoring, 735 
monotonicity, 661 
Monte Carlo, 493 
Morris-Lecar neuron, 719 
motion, 801 
motion analysis, 850 
motion detection, 801,880 
motor control, 38 
motor learning, 3 8 
multi-electrode recording, 131 
multi-layer perceptton, 385, 978 
multi-resolution, 773 
multi-scale, 773 
multi-scale processing, 887 
multi-time models, 1050 
multidimensional prediction, 287 
multimodal integration, 908 
multiple models, 591 
multiple-instance learning, 570 
multiplicative updating, 696 
multiresolution methods, 843 
multiunit recording, 222 
music processing, 887 
mutual information, 794 
mutual source, 465 
naming latency, 59 
natural gradient, 3 85 
natural language processing, 409 
natural scenes, 423,836 
natural stimuli, 103 
natural vision, 236 
navigation, 145, 822 
nearest neighbor, 245, 507 
nearest neighbor algorithm, 612 
neural coding, 110, 131,201 
neural development, 152, 159 
neural network, 808, 908, 957 
neural networks, 598, 661,894 
neural oscillations, 3, 719 
neuro forecasting, 992 
neurological disorders, 66 
neuromorphic, 712 
neuromo/-phic engineering, 880 
neuromorphic systems, 822 
neuron, 124 
neuropsychology, 17, 24, 66 
neuroscience, 24, 103 
noise, 992 
noisy problems, 437 
non-linear, 308 
non-linear analysis, 357 
non-linear principal components analysis, 472 
non-paramelxic, 773 
nonconvex, 350 
nonlinear dimensionality reduction, 682 
nonlinear regression, 535 
nonnegative, 350 
nonparametric estimation, 308 
nonparametric learning, 1008 
nonparametric models, 1008 
nonstationarity, 735, 1057 
numerical methods, 1036 
nystagmus, 138 
object recognition, 17, 215, 829 
object representation, 215 
object-centered, 24 
OCR, 94 
olfaction, 166 
online algorithms, 458 
online learning, 322, 357, 364, 451, 514, 908 
online prediction, 287 
optic aphasia, 66 
optic teeturn, 152 
optimal control, 1029, 1036 
optimal rule, 322 
Keyword Index 1087 
order parameters, 301 
order-parameter dynamics, 357 
orientation detection, 873 
orientation selective, 726 
oscillations, 329 
ovarian cancer, 978 
overcomplete codes, 556, 815 
PAC learning, 336 
packet video and data traffic, 915 
PaCTs, 542 
palindromes, 87 
parallel algorithm, 808 
parameter esfmation, 444 
partial least-squares, 633 
partially observable Markov decision process, 
1015 
path integration, 145 
pattern completion, 654 
pattern manifold, 350, 654 
pattern recognifon, 640, 661,950 
pattern separation, 73 
perceptron decision tree, 336 
perceptual integration, 187 
performance analysis, 978 
periodic attractors, 3 
perspective geometry, 780 
perturbative m-sequences, 180 
phase, 726 
phase transitions, 430 
phonefcs, 3 8 
physical interconnections, 705 
piecewise linear sigmoids, 535 
piecewise-linear models, 661 
place cells, 145 
Poisson process, 110 
policy iteration, 936 
polynomial networks, 1022 
polynomial size, 252 
portfolio management, 93 6 
posterior approximation, 395, 402 
predictability, 315 
prediction, 735, 887 
prediction uncertainty, :36 
prediction with expert advice, 364 
predisposition, 31 
preference function, 451 
principal component analysis, 10, 626, 640 
principal components, 633 
prior information, 661 
prior knowledge, 640 
prioritized sweeping, 1001 
probabilistic inference, 416, 479 
probabilistic transducers, 409 
probability models, 584 
projection pursuit, 273, 423 
pronunciation, 59 
proximity data, 528 
psychophysics, 173,787 
pulse-density-modulation, 705 
Q-learning, 936, 1064, 1071 
quasi-linear, 236 
radial basis functions, 343,402, 829 
radio signals, 756 
rainfall processes, 985 
random design, 294 
rate nmltiplier, 705 
RBF networks, 577 
RCC, 619 
reactive, 1043 
reading, 59 
reading models, 94 
real-time, 957 
receptive fields, 124, 208, 423 
reciprocal pathways, 675 
recognition, 887 
recollection, 73 
rectification, 350 
rectified Gaussian, 486 
rectarent networks, 87, 315, 619, 654, 971 
redundancy reduction, 131 
refractory period, 110 
regime switching, 735 
regression, 493 
regularization, 294, 458 
regularization networks, 343 
Reichardt detector, 880 
reinforcement learning, 145, 922, 936, 1001, 1008, 
1022, 1029, 1036, 1043, 1050, 1057, 
1064, 1071 
relative entropy, 287 
relaxation time, 315 
relevance, 859 
reproductility, 110 
residual algorithms, 1022 
resonance, 138 
retinal ganglion cell, 110 
reversible-jump, 577 
rhythmic expectation, 3 
ridge regression, 294 
risk adjustment, 936 
risk minimization, 343 
robust estima6on, 808, 843 
robust statistics, 80 
ROC-curves, 978 
role based, 957 
saddle point, 329 
scheduling, 1057 
seasonal variation, 985 
second order motion, 801 
segmentation, 80, 850 
selective attention, 859 
self-organization, 549 
self-organizing map, 430, 486 
1088 Keyword Index 
self-similarity, 836 
sensorimotor adaptation, 38 
sensory coding, 836 
sensory feedback, 38 
sequential experiment design, 528 
sequential learning, 458 
Shannon limit, 479 
shape-from-shading, 787 
signal detection, 201 
signal estimation, 201 
silicon retina, 712, 873 
similarity measures, 465 
simple cell, 801 
simulation-based optimization, 1022 
single-trial, 901 
singular value decomposition, 45,626 
skew, 423 
small sample size, 437 
soft-committee machine, 357 
source channel coding, 430 
sparse coding, 794, 815 
spatial frequency, 17 
spatial representation, 24 
speech perception, 38 
speech production, 3 8 
speech recognition, 749, 763 
spike generator, I I 0 
spike sorting, 222 
spike timing, 166, 187 
spiking neurons, 194 
stacking, 668 
statistical image model, 773 
statistical independence, 563 
statistical mechanics, 301,322, 357, 378 
statistical mixture models, 117 
stereo disparity, 486 
stereo vision, 808 
stimulus associabilities, 117 
stochastic dynamics, 1029 
stochastic optimization, 301, 444 
stock prediction, 570 
storage capacity, 378 
straight-line code, 929 
striate cortex, 801 
structural risk, 308 
structure learning, 521 
structure removal, 423 
substitution PACT, 542 
sufficient statistics, 794 
suffix trees, 409 
superadditivity, 66 
supervised learning, 929 
support vector machines, 343, 507, 640 
surface perception, 787 
SVD, 45 
symbol-sensitive counting, 87 
synapses, 194 
synaptic depression, 194 
synaptic transmission, 201 
synchronization, 187 
tangent distance, 647, 843 
telecommunications, 922 
template matching, 829 
temporal, 712 
temporal abstraction, 1050 
temporal differences, 145 
test eror estimation, 437 
test inputs, 437 
tetrodes, 222 
text, 45 
texture, 866 
texture synthesis, 773 
textures of textures, 866 
theory, 252 
threshold element, 252 
time constant, 705 
time series, 308, 315, 735, 985 
top-down, 605 
top-down expectations, 80 
topographic mapping, 549 
topography, 486 
tracking, 908 
traning, 371 
transducers, 409 
trends, 985 
tuning, 124 
turbocodes, 479 
turbulent flows, 836 
unconscious inference, 605 
uniform convergence, 308 
unsupervised learning, 131,528, 556, 605, 654, 
668, 815, 887, 901 
up-down counter, 705 
up-propagation, 605 
user modeling, 943 
value of information, 528 
variational bounds, 416 
vario-eta, 992 
VC-dimension, 259 
vector quantization, 430 
ventral tegmental area, 117 
vesicle release, 194 
vestibulo-ocular reflex, 138 
hew-tuning, 215 
vasion, 712, 787, 794, 859 
vtsion chip, 873, 880 
visual attention, I0 
visual cortex, 80, 208 
visual recognition, 80 
visual search, I0 
visual structure, 866 
VLSI, 719, 726 
wavelets, 915 
Keyword Index 1089 
weight space structure, 378 
weighted regression, 633 
winner-take-all, 215 
Winnow, 500 
word/non-word recognition, 94 
