File Exchange

image thumbnail

Deep Learning Network Analyzer for Neural Network Toolbox

by MathWorks Deep Learning Toolbox Team
Visualize and Analyze Deep Learning Networks
5.0
15 Ratings

23 Downloads

Updated 18 Apr 2018

Network Analyzer analyzes the deep learning network architecture specified by layers. The layer information includes the sizes of layer activations and learnable parameters, the total number of learnable parameters, and the sizes of state parameters of recurrent layers. Use the network analyzer to
• Visualize and understand the network architecture,
• Check that you have defined the architecture correctly,
• Detect problems before training.
Problems that analyzeNetwork detects include
• Missing or disconnected layers,
• Mismatching or incorrect sizes of layer inputs,
• Incorrect number of layer inputs,
• Invalid graph structures.
Usage Example:
net = googlenet
analyzeNetwork(net)

Comments and Ratings (21)

boris fain

Hi - apologies - I'm on 2018a and would like to download this but the only thing that comes down is the network_analyzer.mlpkginstall file which contains instructions to open it (matlab doesn't want to) and a long software license in an xml format. How do I download this?

Seunghoon Lee

ming zhu

Venkat

Hi, Will this work only for DAG or will it work for Series network as well? I have un-pooling layers in my series network and the analyzer always shows error as if the index and size output from pooling layers are not connected to the respective un-pooling layers. When I plot the DAG network, it shows they are connected. Kindly help

gang qin

How is this network used? How to retrain? So input and output data? Which god can help to explain it?

gang qin

Vineet Ahirkar

Arie Lerner

Nice, How do I get a total number of network learnables? I can't copy the list to excel, need to sum them manually :)

Nick Durkee

Currently on 2018b and I can't download this toolbox nor can I use analyzeNetwork command as it's supposed to be implemented in the prerelease. I have the deep learning toolbox.

Daniel

As of 2018b, this appears to be integrated into the Deep Learning Toolbox (formerly the Neural Network Toolbox). Huzzah!

Akio Murakami

Fangqi Zhu

Daniel

Well, for starters I'd just like to see this *exactly as it is* incorporated into Matlab, because it's great. But if you want a wish list, it would be great to incorporate the generation of the horizontal layer diagrams that are typically incorporated into research papers, e.g. this one I chose randomly from Google Images: https://bit.ly/2KYvClC

Maria Duarte Rosa

Hi @Daniel, thank you for your comment. Can you tell us more exactly what sort of convnet diagrams you would like to see in the Network Analyzer? Do you have a specific example? We appreciate your feedback.

Daniel

This is fantastic... is this on the roadmap for integrating into the NN toolbox? This doesn't feel like an extra, it feels like an absolute necessity. Also, are there plans to incorporate convnet diagrams into this toolkit?

Thanks!

Nickolas LaSorte

Amod Anandkumar

Anand

Tarique Ahmad

Very Helpful for diagnostic purposes

adel adel

awesome!!

Mahmut Okatan

Thanks a lot! It would be awesome if analyse result could be exported as well e.g. as a cell array.

MATLAB Release Compatibility
Created with R2018a
Compatible with R2018a
Platform Compatibility
Windows macOS Linux
Acknowledgements

Inspired: Pre-trained 3D ResNet-18

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!