0.25
0.5
0.75
1.25
1.5
1.75
2
Bayesian Neural Nets
Published on Oct 11, 20187379 Views
Related categories
Chapter list
Bayesian Deep Learning00:00
Model Selection00:58
Bayesian or Frequentist?03:57
Bayesian Deep Learning05:50
How do we build models that learn and generalize? - 109:35
How do we build models that learn and generalize? - 210:06
How do we build models that learn and generalize? - 311:13
How do we build models that learn and generalize? - 412:59
Bayesian Inference14:02
Predictive Distribution14:39
Example: Bent Coin - 116:36
Example: Bent Coin - 218:42
Example: Bent Coin - 319:59
Beta Distribution20:36
Example: Bent Coin - 421:32
A Function Space View: Gaussian processes23:18
Gaussian processes24:43
Linear Basis Models - 126:55
Linear Basis Models - 227:43
Linear Basis Function Models27:54
Example: RBF Kernel28:45
Sampling from a GP with an RBF Kernel - 131:41
Sampling from a GP with an RBF Kernel - 232:21
RBF Kernel Covariance Matrix33:06
Learning and Predictions with Gaussian Processes33:45
Inference using an RBF kernel - 135:24
Inference using an RBF kernel - 237:13
Learning and Model Selection - 137:53
Learning and Model Selection - 240:21
Aside: How Do We Build Models that Generalize?40:42
Gaussian Processes and Neural Networks42:59
Deep Kernel Learning - 145:42
Deep Kernel Learning - 246:47
Scalable Gaussian Processes47:21
Deep Kernel Learning Results49:37
Face Orientation Extraction50:17
Learning Flexible Non-Euclidean Similarity Metrics51:01
Step Function51:49
LSTM Kernels52:56
GP-LSTM Predictive Distributions53:10
The Bayesian GAN54:04
Wide Optima Generalize Better56:37
Loss Surfaces in Deep Learning59:06
Mode Connectivity01:01:09
Example Parametrizations01:02:09
Connection Procedure with Tractable Loss01:02:44
Curve Ensembling01:03:27
Fast Geometric Ensembling01:05:59
Trajectory of SGD - 101:06:44
Trajectory of SGD - 201:07:04
Trajectory of SGD - 301:07:06
Trajectory of SGD - 401:07:18
Trajectory of SGD - 501:08:14
Following Random Paths01:08:43
Path from wSWA to wSGD01:09:14
Approximating an FGE Ensemble01:10:21
SWA Results, CIFAR01:10:52
SWA Results, ImageNet (Top-1 Error Rate)01:11:40
Sampling from a High Dimensional Gaussian01:12:09
High Constant LR01:13:27
Conclusions01:14:49
Deriving the RBF Kernel01:21:10