INDEX
AlexNet, 135–143, 148, 152, 157, 158, 161, 165. See also Neural networks
AlphaGo, 3, 7, 228–248, 257, 259, 263, 265. See also Reinforcement learning
Arcade Learning Environment (ALE), 99, 121, 227, 252, 259, 265. See also Reinforcement learning
Atari, 3, 89–123, 135, 137, 146, 224–227, 234, 236, 237, 242, 247, 252, 257–260, 265. See also Reinforcement learning
Automata, 1–6, 61, 92, 126, 128, 134, 168, 180, 207, 248, 261–262, 266
Blending, 73, 74, 80–86, 150, 203
Boosting, 81–82. See also Blending
Boss (self-driving car), 37–55, 119, 160, 164, 182, 184, 224. See also Self-driving cars
Branching ratio, 214, 217, 223, 224, 226, 228, 234, 251
Carnegie Mellon University (CMU), 9, 10, 24, 36, 36, 223, 224
Churchill, David, 251, 253, 259
Classifiers (overview), 59–62
Controllers, 13–14, 26, 35, 39, 52–54, 252
DARPA. See Defense Advanced Research Projects Agency (DARPA)
DeepMind, 89–92, 99–100, 105, 107, 111–112, 121, 146, 224–226, 230, 234, 236–238, 242, 246–248, 259, 263
DeepQA. See Watson (IBM), DeepQA
Deep Blue, 7, 91, 175, 220–225, 228, 245–246, 262–264
Deep Learning. See Neural networks
Deep Speech II, 158, 162. See also Neural networks
Defense Advanced Research Projects Agency (DARPA), 20–24, 36–42, 56–57, 87, 171, 263
de Vaucanson, Jacques, 1–6, 11, 15, 59, 126–129, 266
DOTA II, 256–258
Ensembles (machine learning). See Blending
Evaluation functions, 219–228, 231, 233, 241–243, 246–247, 251, 263
Fan, James, 171, 176, 187–188, 192, 229. See also Watson (IBM), history
Fei-fei Li, 133–135. See also ImageNet
Feng-hsiung Hsu, 223–225, 245. See also Deep Blue
Ferrucci, David, 176–178, 187, 202, 204. See also Watson (IBM), history
Finite state machines, 39, 42–44, 52, 54, 91, 160, 262
deep learning, 142, 144–146, 152–153, 158, 167, 168, 202
DeepMind, 89, 91, 107, 230, 239
Fei-Fei Li, 133–134
self-driving cars, 51, 54, 89, 91
Hassabis, Demis, 89, 107, 259. See also DeepMind
IBM
computer Go, 230
Deep Blue, 91, 220, 222–226, 229, 262, 264 (see also Deep Blue)
Watson, 7, 169, 172–173, 175–177, 187, 190, 198, 203, 205, 263, 266 (see also Watson)
ImageNet, 133–136, 139–142, 151, 158, 265
Inception Network, 142, 202. See also Neural networks
Jeopardy, 3, 6, 7, 91, 169, 171–206, 226, 229, 262, 265. See also Watson
Kalman filters, 18–19, 31, 40–41
Koren, Yehuda, 80. See also Netflix Prize
Language model, 163–164
Lidar, 19, 21, 25–31, 35, 39, 40, 180
Matrix factorization, 67–71, 74–76, 80, 81, 87, 108
MCTS. See Monte Carlo Tree Search
Mechanical Turk, 126–128, 134, 208
Monopoly board, 39, 42–49, 52, 91, 160, 184, 253. See also Finite state machines; Self-driving cars
Montezuma’s Revenge (game), 122, 226, 258
Monte Carlo Tree Search, 241–245. See also AlphaGo
Natural language processing, 178, 181, 203, 205. See also Sentence parsing
Netflix Prize, 57–87, 108, 131, 135, 150, 199, 203, 205, 254, 263, 264
Netflix Prize teams
BellKor, 57, 58, 64–65, 70, 72–75, 77–86
BellKor’s Pragmatic Chaos, 85–86
Dinosaur Planet, 58, 72, 74, 80, 82, 85 (see also Netflix Prize)
The Ensemble, 86
Gravity, 58, 72, 74, 80, 82, 83, 84, 85
Pragmatic Theory, 58, 78, 83–85, 205, 254, 264 (see also Netflix Prize)
Neural networks
activation (squashing) functions, 117, 119, 146–149
artificial neurons, 109–118, 139–156
caption generation, 141, 164–168, 258
convolutional, 115–123, 135–143, 150–151, 156–166, 236–237, 246–247, 265
deep learning, 10, 55, 125, 139, 142–144,145, 203
Filter (see Neural networks, convolutional)
long short-term memory (LSTM), 166–168
optical illusions, 151
rectified Linear Unit (ReLU), 149–150
Recurrent Neural Network (RNN), 159–162, 164–167, 257–258
Occam’s razor, 131–133
Off-policy learning, 100, 260. See also Reinforcement learning
Overfitting, 79–80, 131–133, 142. See also Classifiers
Parsing. See Sentence parsing
Perception, 6–7, 144, 253, 262, 266
in Atari games, 92, 99, 113, 119, 121 (see also Neural networks, convolutional)
auditory (see Speech recognition)
in self-driving cars, 23, 25–27, 32–33, 39–42, 49–55, 59
visual, 107, 144, 157, 160 (see also Neural networks, convolutional)
in Watson, 180
Pruning, 212–214, 217–218, 221–223, 228, 231, 233
Real-time strategy games, 250. See also StarCraft
Red Team (self-driving cars), 9, 14, 16, 18–24, 27, 29, 30, 33, 46. See also Self-driving cars
Reinforcement learning. See also Off-policy learning; Temporal discounting
for Atari games, 91–103, 114, 262
golf (example), 93–119
with neural networks, 108, 129, 257
TD-Gammon, 226–227
Self-driving cars
architectures, 25, 32–33, 37, 39, 42, 50–54, 119, 252
DARPA Grand Challenge (first), 9–22
DARPA Grand Challenge (second), 23–36
DARPA Urban Challenge, 37–56
Neural networks, 10–11, 59 (see also Perception, visual)
path search, 15–18, 33, 39, 42, 45–46, 48, 50, 52
perception in (see Perception, in self-driving cars)
Sentence parsing, 181–185, 196, 199. See also Watson, question analysis
Sequencers, 39, 52–54, 253. See also Monopoly board
Singular extensions, 223–224, 245–246
Speech recognition, 157–165, 209
Stanford Racing Team, 23–30, 33–38, 50, 59, 263, 265
Stanford University, 24, 59, 133, 263
Stanley (self-driving car), 33–35, 38, 40, 42, 47, 50–51, 59, 119, 171, 262. See also Self-driving cars
Sudoku, 208–214, 217, 218, 225
Temporal discounting, 97–99, 103, 108, 114, 257. See also Reinforcement learning
Three-layer architecture, 39, 50–54, 58, 253. See also Self-driving cars, architectures
Thrun, Sebastian, 24–25, 28–31, 56, 59, 171, 262–265
Turk. See Mechanical Turk
University of Alberta, 99, 121, 229, 249, 254, 259, 263
Urmson, Chris, 9–10, 14–15, 18–22, 36–38, 49–51, 55–56. See also Self-driving cars
Watson (IBM), 3, 7, 91, 169, 171–206, 218, 226–230, 266
candidate generation, 189–193
DeepQA, 177–178, 184, 185, 188, 201, 203, 205
evidence retrieval, 194–197
history, 171–172, 175–177, 187–188
question analysis, 178–185 (see also Sentence parsing)
ranking, 199–202
scoring, 197–199