en-es
en-fr
en-sl
en
0.25
0.5
0.75
1.25
1.5
1.75
2
Marrying Graphical Models & Deep Learning
Published on Jul 27, 20178234 Views
Related categories
Chapter list
Marrying Graphical Models & Deep Learning00:00
Overview:00:15
ML as Statistics01:44
Bias Variance Tradeoff05:14
Graphical Models09:18
Bayes ball algorithm13:02
Markov Random Fields14:34
Latent Variable Models15:48
Approximate Inference17:18
Independence Samplers & MCMC20:22
Sampling 101 – What is MCMC?22:49
Sampling 101 – Metropolis-Hastings27:26
Approximate MCMC30:45
Minimizing Risk32:12
Stochastic Gradient Langevin Dynamics34:47
Demo: Stochastic Gradient LD38:37
A Closer Look … - 139:20
A Closer Look … - 240:35
Demo SGLD: large stepsize41:15
Demo SGLD: small stepsize41:48
Variational Inference42:44
Learning: Expectation Maximization44:27
Amortized Inference47:45
Deep NN as a glorified conditional distribution50:01
The “Deepify” Operator50:46
Variational Autoencoder52:28
Deep Generative Model: The Variational Auto-Encoder52:39
Stochastic Variational Bayesian Inference53:07
Reducing the Variance: The Reparametrization Trick55:24
Semi-Supervised VAE I - 158:12
Semi-Supervised VAE I - 258:22
Discriminative or Generative?01:03:51
Big N vs. Small N?01:04:02
Combining Generative and Discriminative Models01:05:26
Deep Convolutional Networks01:07:02
So..., CNNs work really well01:07:17
Reasons for Bayesian Deep Learning01:08:24
Example01:09:30
Bayesian Learning01:09:32
Variational Bayes01:12:53
Sparsifying & Compressing CNNs01:13:28
Full Bayesian Deep Learning01:14:48
Stochastic Variational Bayes01:16:16
Local Reparametrization01:16:21
Two Layers01:18:06
Variational Dropout01:18:24
Sparsity Inducing Priors01:18:26
Variational Dropout - 101:18:28
Variational Dropout - 201:19:14
Node (instead of Weight) Sparsification01:19:25
Preliminary Results01:19:26
Conclusions01:20:36