Sign inSign up
Building Machines that Imagine and Reason: Principles and Applications of Deep Generative Models thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Building Machines that Imagine and Reason: Principles and Applications of Deep Generative Models

Published on Aug 23, 201623371 Views

Shakir Mohamed

Deep generative models provide a solution to the problem of unsupervised learning, in which a machine learning system is required to discover the structure hidden within unlabelled data streams. Becau

Deep Learning Summer School 2016 - Montreal

Related categories

Deep LearningReinforcement LearningUnsupervised Learning

Chapter list

Building Machines that Imagine and Reason - Principles and Applications of Deep Generative Models00:00
Motivations for machine learning00:36
Why Generative Models/101:50
Why Generative Models/203:02
Why Generative Models/303:50
Landscape of Generative Models04:45
Data imputation | In-painting | Denoising04:50
Semi-supervised Classification05:35
Communication and Compression06:08
3D Scene Generation06:35
Rapid Scene Understanding07:23
One-shot Generalisation08:13
Environment Simulation08:37
Representation Learning for Control09:01
Visual Concept Learning09:35
Density-based Exploration10:14
Macro-actions and Planning10:57
Successful Applications of Generative Models11:43
Progress in Generative Models/112:12
Progress in Generative Models/212:59
Machine Learning Framework13:14
Types of Generative Models13:46
Smorgasbord of Learning Principles14:14
Combining Models and Inference14:30
A Model for Every Occasion14:44
Types of Generative Models14:52
Fully-observed Models/115:59
Fully-observed Models/217:16
Model-space Visualisation18:45
Transformation Models/119:15
Transformation Models/220:25
Model-space Visualisation24:02
Latent Variable Models/124:34
Latent Variable Models/225:17
Model-space Visualisation27:48
Inference and Learning28:19
Inferential Problems28:37
Bayesian Model Evidence29:18
Importance Sampling30:14
Importance Sampling to Variational Inference31:22
Variational Free Energy32:09
Other Families of Variational Bounds32:52
Bayesian Two-sample Testing/133:50
Bayesian Two-sample Testing/235:11
Testing to Adversarial Learning37:20
Tools for Algorithm Building40:29
Variational EM41:28
Stochastic Approximation42:37
Memoryless Inference44:14
Amortised Inference44:44
Amortised Variational Inference45:49
Minimum Description Length47:02
Amortised Message Passing47:53
Amortised Predictive Distributions48:43
Stochastic Optimisation49:15
Stochastic Gradient Estimators54:33
The Case of Variational Autoencoders55:22
Variational Auto-encoders in General55:32
Implementing a Variational Algorithm56:38
Latent Gaussian VAE/158:44
Latent Gaussian VAE/259:57
VAE Representations01:02:54
Latent Gaussian VAE/301:05:23
Latent Binary VAE/101:06:13
Latent Binary VAE/201:06:50
Semi-supervised VAE01:07:10
Sequential Latent Gaussian VAE/101:07:54
Sequential Latent Gaussian VAE/201:08:34
Sequential Latent Gaussian VAE/301:09:47
Sequential Latent Gaussian VAE/401:09:54
Structured Sequential VAEs/101:10:23
Structured Sequential VAEs/201:11:34
Volumetric VAEs/101:12:13
Volumetric VAEs/201:13:06
Macro-action Learning/101:16:53
Macro-action Learning/201:17:14
Summary - Demonstrated Applications of Generative Models01:17:17
Summary/101:17:30
Summary/201:17:49
Summary/301:18:02
The Future of Generative Models01:18:15
Thank You!01:18:58