File:Infinitely wide neural network.webm

From Vigyanwiki

Original file(WebM audio/video file, VP8/Opus, length 8.0 s, 1,920 × 1,080 pixels, 9.87 Mbps overall)

This file is from Wikimedia Commons and may be used by other projects. The description on its file description page there is shown below.

Summary

Description
English: Left: a Bayesian neural network with two hidden layers, transforming a 3-dimensional input (bottom) into a two-dimensional output (top). Right: output probability density function induced by the random weights of the network. Video: as the width of the network increases, output distribution simplifies, ultimately converging to a multivariate normal in the infinite width limit. Note: the video is an artistic depiction of the progression, and not an actual simulation.
Date
Source https://ai.googleblog.com/2020/03/fast-and-easy-infinitely-wide-networks.html
Author Tom Small

Presented in https://iclr.cc/virtual_2020/poster_SklD9yrFPS.html

Licensing

w:en:Creative Commons
attribution share alike
This file is licensed under the Creative Commons Attribution-Share Alike 4.0 International licence.
You are free:
  • to share – to copy, distribute and transmit the work
  • to remix – to adapt the work
Under the following conditions:
  • attribution – You must give appropriate credit, provide a link to the licence, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.
  • share alike – If you remix, transform, or build upon the material, you must distribute your contributions under the same or compatible licence as the original.

Captions

Add a one-line explanation of what this file represents
As the network becomes infinitely wide (left), output distribution converges to a Gaussian process (right)

13 March 2020

video/webm

File history

Click on a date/time to view the file as it appeared at that time.

Date/TimeThumbnailDimensionsUserComment
current00:50, 8 August 20208.0 s, 1,920 × 1,080 (9.41 MB)wikimediacommons>NovakRomanUse VP8 encoding instead of VP9

Metadata