Automating stochastic gradient methods with adaptive batch sizes
57 mins 22 secs,
835.49 MB,
MPEG-4 Video
640x360,
29.97 fps,
44100 Hz,
1.94 Mbits/sec
Share this media item:
Embed this media item:
Embed this media item:
About this item
Description: |
Goldstein, T
Wednesday 6th September 2017 - 09:50 to 10:40 |
---|
Created: | 2017-09-07 13:37 |
---|---|
Collection: | Variational methods and effective algorithms for imaging and vision |
Publisher: | Isaac Newton Institute |
Copyright: | Goldstein, T |
Language: | eng (English) |
Distribution: | World (downloadable) |
Explicit content: | No |
Aspect Ratio: | 16:9 |
Screencast: | No |
Bumper: | UCS Default |
Trailer: | UCS Default |
Abstract: | This talk will address several issues related to training neural networks using stochastic gradient methods. First, we'll talk about the difficulties of training in a distributed environment, and present a new method called centralVR for boosting the scalability of training methods. Then, we'll talk about the issue of automating stochastic gradient descent, and show that learning rate selection can be simplified using "Big Batch" strategies that adaptively choose minibatch sizes. |
---|
Available Formats
Format | Quality | Bitrate | Size | |||
---|---|---|---|---|---|---|
MPEG-4 Video * | 640x360 | 1.94 Mbits/sec | 835.49 MB | View | Download | |
WebM | 640x360 | 464.25 kbits/sec | 195.07 MB | View | Download | |
iPod Video | 480x270 | 522.21 kbits/sec | 219.42 MB | View | Download | |
MP3 | 44100 Hz | 249.78 kbits/sec | 105.04 MB | Listen | Download | |
Auto | (Allows browser to choose a format it supports) |