Unreasonable Effectiveness of Learning Artificial Neural Networks thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Unreasonable Effectiveness of Learning Artificial Neural Networks

Published on Nov 28, 20162682 Views

Deep networks are some of the most widely used tools in data science. Learning is in principle a hard problem in these systems, but in practice heuristic algorithms often find solutions with good gene

Related categories

Chapter list

Unreasonable Effectiveness of Learning Artificial Neural Networks00:00
Plan of the talk00:40
Computational neuroscience & artificial neural system01:11
Dissecting Deep Neural Networks - 102:59
Dissecting Deep Neural Networks - 205:05
Dissecting Deep Neural Networks - 306:04
Convolutional layers07:33
Restricted Boltzmann Machines layers08:38
Auto-encoders layers - 109:18
Auto-encoders layers - 209:25
Auto-encoders layers - 310:12
Auto-encoders layers - 411:05
Auto-encoders layers - 512:56
Auto-encoders layers - 613:16
Speech recognition13:52
Deep translator - 114:58
Deep translator - 215:22
Current and short term applications - 116:22
Current and short term applications - 217:44
human vs. algorithms19:48
Search tree for Chess20:09
The game of GO20:56
Search tree for Go21:55
AlphaGo Overview - 122:15
AlphaGo Overview - 223:44
AlphaGo Overview - 324:03
Next challenges25:13
Still far away from AGI26:58
How does learning take place?29:01
Learning ~ energy minimisation problem29:15
The simplest neural device31:21
Geometry of the space of solutions32:53
Golf course34:36
Local entropy measure- 138:29
Local entropy measure - 238:49
Local entropy measure - 539:27
Principled algorithm40:51
Local entropy driven stochastic search41:15
Conclusions42:01