evolution-strategies-starter
evolution-strategies-starter copied to clipboard
openai
Code for the paper "Evolution Strategies as a Scalable Alternative to Reinforcement Learning"
According to the paper on page 3, Algorithm 2, the gradient in line 11 is rescaled by the standard deviation. However I can't see it in the code in: https://github.com/openai/evolution-strategies-starter/blob/master/es_distributed/es.py#L247...
Obviously very problem dependent as usual, but for the noise of the perturbations this is set to the 0.02 standard dev in the config (also in the paper apparently we...
Hello, why use batched_weighted_sum instead of direct dot on all items?
If I understand the code correctly, it uses virtual batch normalization only for the inputs and **not** for the intermediate layers. Was this done in the Atari context for getting...
tf.concat(concat_dim, values, ) for tensorflow 1.0+, the api tf.concat changed
hi, Thanks for code sharing! after reading the source code, l have such questions, could you help me better understand the code? 1. why design ac_noise? rather than deterministic action...
I've been posting some ideas about evolution around the internet, so why not here! “ Quantization is the enemy of evolution It is fortunate that biological systems are heavily quantized,...
Thank you very much for this great job! I was wondering if there exists a local version of this implementation that doesn't need EC2 to run. It would make it...
Related to #5 but to say that any config that doesn't assume a mujoco license would be great (including AMI-packing step if that is relevant).
Hi, I would like to implement ES on text data can you please provide some intuitions on implementing ES for text classification use-cases. Possibly should I injects noise directly in...
