0.25
0.5
0.75
1.25
1.5
1.75
2
The Neural Autoregressive Distribution Estimator, incl. discussion by Yoshua Bengio
Published on May 06, 20117345 Views
We describe a new approach for modeling the distribution of high-dimensional vectors of discrete variables. This model is inspired by the restricted Boltzmann machine (RBM), which has been shown to be
Related categories
Chapter list
The Neural Autoregressive Distribution Estimator00:00
Distribution Estimation00:07
In this talk...01:24
Restricted Boltzmann Machine02:04
Fully Visible Bayesian Networks03:58
Turning an RBM into a Bayesian Network (1)05:56
Turning an RBM into a Bayesian Network (2)06:27
Turning an RBM into a Bayesian Network (3)07:18
Turning an RBM into a Bayesian Network (4)07:44
Turning an RBM into a Bayesian Network (5)08:11
Turning an RBM into a Bayesian Network (6)08:16
Neural Autoregressive Distribution Estimator (NADE) (1)10:29
Neural Autoregressive Distribution Estimator (NADE) (2)12:12
Neural Autoregressive Distribution Estimator (NADE) (3)13:29
Neural Autoregressive Distribution Estimator (NADE) (4)14:32
Related Work15:50
Experiments (1)18:37
Experiments (2)19:51
Conclusion21:02
Thank you!22:06
Discussant Comments on Larochelle & Murray’s "The Neural Autoregressive Distribution Estimator"22:11
NADE: Neural Autoregressive Distribution Estimator22:30
Left-to-Right directed neural-net belief networks (Bengio & Bengio 2000)23:09
Parametrize as RBM conditional’s meanfield recursion 1st iteration: NADE24:12
Comments on Results24:43
Sharing Ordering26:10
More Discussion27:12
NADE27:59