timoklein / redo Goto Github PK
View Code? Open in Web Editor NEWReDo: The Dormant Neuron Phenomenon in Deep Reinforcement Learning (pytorch)
ReDo: The Dormant Neuron Phenomenon in Deep Reinforcement Learning (pytorch)
Hello,
I saw that you added a line in README:
NOTE: This implementation is not correct as of now. I plan to fix it in the coming week!
Could you please write more about what is not correct? Is it a tiny issue like an edge case or something very big? I am writing some code based on your implementation and will be analyzing it thoroughly, so maybe I will be able to help.
The flat-mask (ex: (32,)
) from dead_neurons gets repeated to make the mask 4D shape of the conv-net ( ex: (8,8,4,32)
) here: https://github.com/google/dopamine/blob/master/dopamine/labs/redo/weight_recyclers.py#L652. does the following initialization have the same effect? : https://github.com/timoklein/redo/blob/main/src/redo.py#L17
In the paper, it says reinitialize their incoming weights and zero out the outgoing weights. I'm confused since in my mind each layer of the network is just a matrix. I'm wondering what are the incoming weights and outgoing weights.
Could you give me some hints?
Thanks a lot!
fc1
has (3136,512) params and it seems like current implementation always resets
and sets 0
to the dead-neuron for out-going layer-512. The implementation is supposed to reset
the dead-neurons of the incoming-layer
and set 0
the dead-neuron of the outgoing-layers
. https://github.com/timoklein/redo/blob/main/src/redo.py#L120
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.