KEYWORD INDEX 
a priori knowledge, 50 
active learning, 483 
Adaline III, 244 
adaptation, 820 
adaptive control, 295,303,319,335 
adaptive filtering, 921 
AIC, 607 
alphabet recognition, 712 
ambiguity resolution, 417 
analog, 789 
Cochlear model, 666 
computation, 666 
simulator (of FitzHugh-Nagumo 
model), 993 
stack, 65 
VLSI, 244, 666,765,773,828,836 
analysis techniques, 27 
annealed approximation, 523 
annealing, 789 
aperiodic dynamics, 556 
applications, 755 
approximate dynamic programming, 319 
approximation power, 615 
area MST, 433 
area MT, 369 
arithmetic functions, 59 
artificial synapse, 813 
associative memory, 34, 236 
asymmetric networks, 556 
asymptotic jackknife estimator, 599 
asymptotic theory, 607 
asymptotics, 599 
asynchronous, 820 
attractor neural networks, 572, 805 
attractor transitions, 236 
attractors, 631 
audition, 820 
auditory cortex, 1007 
autoconfiguration, 797 
autoencoder networks, 279,937 
autonomous navigation, 279 
axon, 781 
balanced error, 73 
basin hopping, 507 
batch training, 459 
Baum-Welch, 11 
Bayes decision rule, 688 
Bayesian, 385 
analysis, 539 
inference, 401 
learning, 11,475 
models, 483 
networks, 393 
updating, 572 
BCM learning, 499 
benchmarking, 739 
best-first search, 11 
bifurcations, 236 
binary weights, 591 
binding problem, 903 
binomial approximations, 564 
biological modeling, 1022 
bistable system, 993 
Boltzmann machine, 172, 531 
boosting, 42 
bootstrap methods, 196 
boundary hunting classifier, 139 
C+ +, 797 
canonical discriminant analysis, 27 
capacity control, 164, 228 
capacity tuning, 147 
CART algorithm, 131 
cascade correlation, 188 
Cauchy basis functions, 871 
cellular automata, 631 
censoring, 572 
cerebellum, 961 
certainty factors, 107 
channel capacity, 515 
channel equalization, 674 
chaos, 549,556,631 
character recognition, 42, 723 
character segmentation, 731 
1043 
1044 Keyword Index 
chess, 204 
circuit depth, 59 
classical conditioning, 937 
climbing fibers, 961 
clustering, 425 
CMAC architectures, 319 
coding theory, 89 
cognitive modeling, 887 
combinatorial optimization, 401 
combinatorial structure, 27 
committee machine, 523 
committee of learners, 483 
competitive learning, 220 
compositional hierarchy, 401 
compositional representations, 855 
conductance-based models, 813 
confidence intervals, 196 
confusion vectors, 895 
connection structure analysis, 515 
consistency, 196,696 
constrained architectures, 863 
constructive algorithm, 115 
constructive learning, 131 
constructive training, 139 
context-dependence, 649 
context-free grammars, 863 
context-free languages, 65,847 
continuous speech recognition, 649, 712 
contour completion, 409 
control, 271 
direct and indirect, 303 
parameters, 549 
correlated activity, 985 
correlation of spike trains, 515 
correlational learning rules, 961 
coupled oscillators, 172 
covariance rule, 499 
crayfish, 993 
cross-validation, 228,475,539,599, 649 
data compression, 580 
data matrixes, 164 
decision tree, 204 
depth planes, 385 
descending control, 921 
descent, 789 
desired signal design, 688 
development, 99,969 
development-environment, 797 
diagnosis, 204, 755 
dichotomy implementation, 639 
diffusion, 459 
digital circuits, 805 
digital programmable circuits, 805 
dimensional separability, 377 
dimensionality reduction, 580,739 
directional units, 172 
discriminants, 499 
distributed representations, 895 
DNA pattern recognition, 204 
dual problems, 147 
duration modeling, 712 
dynamic networks, 688 
dynamic programming, 263,271,295,342, 
723, 731 
ego-motion, 433 
eigenvector, 156 
elastic network, 255,401 
electric fish, 921 
electric power demand, 739 
electrosensory system, 921 
EM, 393 
EMG (electromyography), 658 
encoder, 89 
energy-efficient, 820 
energy maximization, 385 
ensemble dynamics, 451 
entropy, 204, 499 
epileptic spike detection, 688 
equilibrium point control, 335 
equivalence of activation functions, 615 
error functions, 696 
error-correcting codes, 89 
event-address protocol, 820 
evidence, 539 
excitable system, 993 
expert knowledge, 871 
exploration, 263 
eye movement, 433,441 
facilitation, 781 
fault tolerance, 89, 491 
feature detector, 1007 
feature maps, 123 
feature selection, 739 
feature space partitioning method, 674 
feudal reinforcement, 271 
Keyword Index 1045 
figure-ground segmentation, 828 
filtering examples, 483 
finite difference method, 212 
first-passage time, 507 
Fisher's method of scoring, 164 
FitzHugh-Nagumo model, 993 
Fokker-Planck equation, 451,507 
forecasting, 739 
formal language learning, 847,863 
forward dynamics model, 658 
foveation maps, 969 
function approximation, 350 
function-counting theorem, 639 
fuzzy logic, 350, 871 
gain control, 921 
game of life, 631 
gating network, 369 
Gaussian mixtures, 139,871 
Gaussian networks, 393 
Gaussian synapses, 765 
gene doubling, 99 
generalization, 164, 188,342,491,523,607 
generalized portrait, 147 
generative, 99 
genetic algorithms, 99 
geometry learning, 887 
gesture recognition, 393 
Gibbs energy model, 531 
gradient, 156,212, 789 
grammar induction, 11, 65, 99, 180,236, 
847, 863 
grasping, 311 
growing network, 123 
Hamilton-Jacobi-Bellman equation, 319 
Hamiltonian dynamics, 475 
hamming network, 564 
hand gesture recognition, 393 
handwritten character recognition, 50,441 
hardware implementation, 212, 773 
harmonic analysis, 19 
harmony, 847 
heading direction, 433 
Hessian, 156,164 
hidden Markov model, 11,649,723,731 
hidden unit space, 27 
hierarchy, 115,271 
higher-order network, 115,342 
hints, 73 
hippocampus, 556,929,937, 1030 
history-dependent dynamics, 572 
holographic memory, 34 
Hopf bifurcation, 1001 
Hybrid Monte Carlo, 475 
hybrid network, 3,649,723,813 
hyperplane movement, 204 
identity map, 311 
image alignment, 731 
image processing, 425,580,755,828 
image segmentation, 172,425,441 
implementation, 789 
.ncremental learning, 11, 115 
infomax principle, 953 
information compression, 311 
information gain, 499 
nformation theory, 350,515,953, 1030 
nput resistance, 921 
nsect, 1022 
ntegrate-and-fire neural network, 1001 
ntegration, 311 
ntersecting regions, 27 
nvariance, 73 
nvariant classification, 50 
nvariant distance, 50 
nvariant object recognition, 903 
nverse kinematics, 255 
inverse model controller, 303 
iterated addition, 59 
K-means clustering, 139 
KBANN algorithm, 887 
Kohonen maps, 123 
Kolmogorov equation, 451,507 
language modeling, 847,895 
lateral excitation, 417 
lateral geniculate nucleus, 409 
lattice networks, 631 
learning, 73,188,789 
adaptive, 467 
unsupervised, 417 
algorithm, 73,499 
control, 327,335 
curve, 607 
dynamics, 631 
from examples, 73 
1046 Keyword Index 
queries, 483 
rate maximization, 156 
schedules, 73 
speed, 204 
trajectory, 491 
vector quantization (LVQ), 220 
letter recognition, 712 
lexical representations, 895 
likelihood ratio, 572 
limited precision, 212 
linear quadratic regulation, 295 
linear threshold functions, 59 
Linsker networks, 1007 
LMS tree, 335 
lobster, 913 
local inhibition-local feedback, 805 
local learning rules, 961 
local linear map (LLM), 188 
local minima, 459 
location-based attention, 903 
low firing, 572 
low-bandwidth, 820 
low-power, 820 
lower bounds, 615 
on network-size, 19 
Mackey-glass, 188 
MAP estimate, 539 
Markov random field (MRF), 385 
MARS algorithm, 131 
matched filter, 688 
matching network, 401 
maximum information preservation, 953 
maximum margin classifier, 147 
maze learning, 271 
mean field theory, 172 
mechanoreceptor cell, 993 
medical diagnosis, 755 
membership queries, 483 
membrane conductance, 921 
memory-based classification, 50 
memory capacity, 180 
memory-based learning, 263 
metamorphosis, 131 
metastability, 523 
Metropolis algorithm, 475 
minimax algorithms, 147 
minimum description length (MDL), 11 
missing features, 393 
mixtures of experts, 369 
model merging, 11 
model selection, 196,228,599,607 
model-free learning, 244,836 
modular architecture, 712 
Monte Carlo integration, 475,871 
mossy fiber, 961 
motion, 781 
detection, 369 
perception, 433,781 
segmentation, 369 
unsmearing, 417 
motor 
control, 311,658, 1014 
learning, 335 
maps, 1014 
MRBF network, 131 
MS-TDNN, 696,712 
multi-cell, 377 
multiple inheritance, 797 
multiresolution filters, 953 
multiresolution speech processing, 666 
muscle model, 913 
musculoskeletal system, 913 
mutual information, 361,515,953 
navigation, 342, 929 
nearest neighbor, 50 
network capacity, 639 
network dynamics, 556 
network interpretation, 863 
network simplification, 164 
neural computation, 1014 
neural development, 985 
neural dynamics, 236 
neural network pushdown automata, 65 
neural network visualization, 204 
neurocontrol, 342 
neuromodulators, 969 
neuron, 1030 
neuron simulator, 813 
neuron-computer interaction, 813 
Newton's method teacher, 303 
N-grams, 895 
node perturbation, 212 
noise, 89,993 
noisy features, 393 
non-linear dynamics, 549 
nonparametric procedure, 196 
Keyword Index 1047 
object recognition, 361,401 
object-based attention, 903 
objective function, 712 
object-oriented simulator, 797 
obstacle avoidance, 342 
ocular dominance, 985 
ocular saccades, 1014 
oculomotor system, 961 
olfaction, 1022 
on-line 
training, 156 
character recognition, 723 
estimation, 319 
ontogenesis, 99,985 
optic flow, 433 
optical character recognition (OCR), 441,731 
optimal brain damage, 164 
optimal brain surgeon, 164 
optimal control, 295,319 
optimal encoding, 953 
optimal equalizer design, 674 
optimal-size neural networks, 19 
optimization, 244, 789 
oscillation, 236,913, 1022 
oscillatory associative memory, 236 
oscillatory automation, 236 
outlier, 467 
overfitting, 188 
PAC learning, 591 
parallel implementation, 836 
PARCOR parameters, 658 
parity, 19 
partition of unity, 765 
pattern recognition, 180,441,704 
peg-in-hole insertion, 327 
penalty terms, 73,228,491,871 
perceptrons, 591 
perceptual organization, 903 
perturbative learning, 836 
phase coding, 929 
phase transition, 523 
phase-locked oscillations, 903 
pheromone, 1022 
phonetic modeling, 649 
place cells, 929, 1030 
planning, 342 
pole balancing, 81 
polynomial networks, 99 
population coding, 377, 1030 
posterior probability, 539 
practice strategy, 335 
predictability minimization, 499 
predicted squared error, 139 
prediction, 188,483 
predictive distribution, 196 
prehension, 311 
priming, 879 
principal components, 467,580 
prior knowledge, 65, 73 
prior probability, 539 
probability estimation, 123 
projection pursuit, 3,739 
pruning, 228,871 
pulse-computation, 781 
pulsed neural networks, 773 
pushdown automata, 65 
Q-learning, 81,263,271,295 
radial basis function, 19, 81,131,139, 147, 
755,765,871 
RAN algorithm, 81, 131 
random connectivity, 549 
random neural network, 1001 
rat navigation, 929 
rational approximation, 19 
rational functions, 623 
read-once, 591 
realizations, 623 
real-time learning, 303 
receptive field, 377, 953,985 
recognition, 311,441 
recognition-based segmentation, 723 
recurrent network, 34, 65, 115,180,244, 
556, 1014 
redundant manipulators, 255 
regularization, 147, 164, 228,255 
Reichardt, 781 
reinforcement learning, 81,263,271,295, 
327,969 
relaxation computation, 311 
reliability estimation, 279 
remote-sensing, 425 
repetition blindness, 879 
replica method, 523 
resampling, 196 
response pattern, 1022 
1048 Keyword Index 
robot, 255,335 
control, 279 
navigation, 263 
path planning, 342 
rule-based, 350,871 
rule-based network, 425,887 
rule refinement, 871 
run-on handwriting, 723 
saccade, 441 
sample reuse, 196 
scale space, 401 
second derivatives, 156 
segmentation, 385,441,723,903 
segmentation-driven training, 723 
segregated maps, 985 
seismic database, 139 
selective sampling, 483 
self-motion, 433 
self-organization, 123,417,467,969, 1007 
semantic representations, 895 
sensory processing, 921 
sensory-motor systems, 342 
sequence learning, 115 
sequence processing, 863 
sequence recognition, 236 
shunting inhibition, 556 
Sigma-pi units, 147, 631 
singular value decomposition (SVD), 895 
skill acquisition, 327 
smooth pursuit, 961 
smoothing, 649 
sparse coding, 953 
sparse networks, 556,572 
spatial frequencies, multiple, 409 
spatial memory, 929 
spario-temporal dynamics, 1001 
speaker independence, 712 
speech 
analysis, 666 
processing, 666 
production, 658 
recognition, 11,204, 649,696,704, 
712,773 
spectrograph, 666 
synthesis, 658 
speed of convergence, 196 
spiral waves, 1001 
stack memory, 65 
statistical mechanics, 467, 523 
stereo vision, 385 
stochastic approximation, 220,459, 467, 836 
stochastic dynamics, 475 
stochastic gradient descent, 156,244 
stochastic learning, 451,507 
stochastic resonance, 993 
stochastic visual grammar, 401 
stomatogastric ganglion, 813,913 
stopped training, 228 
storage capacity, 564 
strabismus, 985 
structural alignment, 855 
structure processing, 863 
structured representations, 855 
superior colliculus, 969, 1014 
symbol manipulation, 847, 863 
synapse, 515 
synaptic activity, 921 
synaptic weight noise, 491 
synchronization, 236 
tangent distance, 50 
temporal coherence, 361 
temporal difference, 81,263,969 
texture recognition, 425 
theory revision, 107 
thesaurus induction, 895 
threshold circuits, 19, 59 
threshold hamming network, 564 
time delay neural network (or TDNN), 417, 
696,712,723 
time series prediction, 580,739 
time warping, 180 
time-delay learning, 115 
topographic maps, 985 
topology-preserving map, 255 
trajectory classification, 180 
trajectory generation, 34 
trajectory learning, 244 
trajectory-specific motion tracking, 417 
transient detection, 688 
translation invariance, 361 
truck backer-upper, 350 
universal approximator, 3 
unsupervised learning, 361,499, 953 
v1,377 
Keyword Index 1049 
VC-dimension, 147 
vector quantization, 220 
velocity estimation, 369 
vestibulo-ocular reflex (VOR), 961 
vision, 279,361,369,441,828 
vision chips, 820 
visual 
attention, 903 
cortex, 369,377,417,433 
development, 417 
maps, 985 
motion, 433 
pattern recognition, 441 
processing, 409 
psychophysics, 433 
Viterbi algorithm, 731 
VLSI, 789,820 
Voronoi tessellation, 220 
wave propagation, 1001 
wavelets, 666,953 
weight 
elimination, 164 
sharing, 631 
space equilibria, 451 
space probabilities, 451,507 
weight-decay, 475,539 
word recognition, 879 
word representation, 879 
word sense disambiguation, 895 
zero-crossing, 781 
