Sign in to confirm you’re not a bot
This helps protect our community. Learn more

Visualizing and Understanding Deep Neural Networks by Matt Zeiler

Data Council
41K subscribers
288K views 10 years ago
Matthew Zeiler, PhD, Founder and CEO of Clarifai Inc, speaks about large convolutional neural networks. These networks have recently demonstrated impressive object recognition performance making real world applications possible. However, there was no clear understanding of why they perform so well, or how they might be improved. In this talk, Matt covers a novel visualization technique that gives insight into the function of intermediate feature layers and the operation of the overall classifier. Used in a diagnostic role, these visualizations allow us to find model architectures that perform exceedingly well. ABOUT DATA COUNCIL: Data Council (https://www.datacouncil.ai/) is a community and conference series that provides data professionals with the learning and networking opportunities they need to grow their careers. Make sure to subscribe to our channel for more videos, including DC_THURS, our series of live online interviews with leading data professionals from top open source projects and startups. FOLLOW DATA COUNCIL: Twitter:   / datacouncilai   LinkedIn:   / datacouncil-ai   ...more
...more
These chapters are auto-generated

Intro

0:00

Overview

0:37

Convolutional Networks (Lecun et al '89)

1:03

Example

2:15

Neural Network History

2:33

Convnets Show Huge Gains

3:44

Deep Nets vs Monkey vs Human

5:11

Deconvolutional Networks

8:51

Deconvolutional Network (2 layers)

9:37

Reversible Max Pooling

9:54

Reuse Feedforward Switches

10:19

Filters

11:18

Projecting back from Higher Layers

12:04

Visualizations of Higher Layers

12:44

Layer 1: Top-9 Patches

17:34

Top-1

18:38

Translation (Horizontal)

24:54

Scale Invariance

26:23

Lack of Understanding

29:25

Visualizations Help -2% Boost

29:39

ImageNet Classification 2013 Results

32:15

Recent Success

32:55

Implicit Correspondence

34:20

Caltech 256

35:00

Tapping off Features at each Layer

36:36

Summary

37:40
Visualizing and Understanding Deep Neural Networks by Matt Zeiler
2.5KLikes
288,164Views
2015Feb 2
Matthew Zeiler, PhD, Founder and CEO of Clarifai Inc, speaks about large convolutional neural networks. These networks have recently demonstrated impressive object recognition performance making real world applications possible. However, there was no clear understanding of why they perform so well, or how they might be improved. In this talk, Matt covers a novel visualization technique that gives insight into the function of intermediate feature layers and the operation of the overall classifier. Used in a diagnostic role, these visualizations allow us to find model architectures that perform exceedingly well. ABOUT DATA COUNCIL: Data Council (https://www.datacouncil.ai/) is a community and conference series that provides data professionals with the learning and networking opportunities they need to grow their careers. Make sure to subscribe to our channel for more videos, including DC_THURS, our series of live online interviews with leading data professionals from top open source projects and startups. FOLLOW DATA COUNCIL: Twitter:   / datacouncilai   LinkedIn:   / datacouncil-ai  

Follow along using the transcript.

Data Council

41K subscribers