alirezamika / evostra

A fast Evolution Strategy implementation in Python
MIT License
270 stars 46 forks source link

Add distribution (multi-processor support) #1

Open alirezamika opened 7 years ago

alirezamika commented 7 years ago

One of the strengths of evolution strategy algorithm is its linear scalability. Thus it would be great to have this implemented in Evostra.

souravsingh commented 7 years ago

@alirezamika I am interested in working on this. How do I start?

alirezamika commented 7 years ago

I haven't started on working on this yet. We can go with the master-slave approach. Slaves should be able communicate with master over network and solve their own task using the given data from master.

hipoglucido commented 6 years ago

Any progress on this?

LinuxIsCool commented 6 years ago

I have been implementing my own version of ES using keras weights, inspired by https://blog.openai.com/evolution-strategies/ . I am currently working on running parallel on cpu. I will let you know if I have progress. You can follow the work here: https://github.com/LinuxIsCool/756project

alirezamika commented 6 years ago

@hipoglucido @LinuxIsCool I've implemented multiprocessing support which is working fine. Next step is to add distribution support over network.