Closed choltz95 closed 4 years ago
Hello @choltz95! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:
Line 27:22: E226 missing whitespace around arithmetic operator Line 39:34: W504 line break after binary operator Line 40:13: E127 continuation line over-indented for visual indent Line 41:13: E127 continuation line over-indented for visual indent
Line 30:13: E127 continuation line over-indented for visual indent
protocols/decentralizedprotocol.py
:Line 58:37: E226 missing whitespace around arithmetic operator
protocols/singleplayerprotocol.py
:Line 20:22: E127 continuation line over-indented for visual indent Line 53:1: W293 blank line contains whitespace Line 63:37: E226 missing whitespace around arithmetic operator
protocols/utils.py
:Line 19:23: E226 missing whitespace around arithmetic operator Line 19:38: E226 missing whitespace around arithmetic operator Line 105:11: E128 continuation line under-indented for visual indent
Can you add a .pep8speaks.yml
file at the root directory with the following texts?
pycodestyle:
max-line-length: 80
ignore:
- E111 # indentation is not a multiple of four
Hi, started work on a simple decentralized protocol for homogeneous bandits which uniformly selects a single learner each iteration and distributes a complete history of all messages to all players, and players broadcast their actions & feedback.
The implementation is a little rough and I could definitely use your feedback. From the config I create a list of identical learners and list of identical bandits. Instead of learner.play(), I call protcol.play(). To do this, I extended regretminlearner to support a centralized regret callback and to run a single iteration given a message history.
Here is a result for uniform, selfish UCB (broadcast no information), & friendly UCB (broadcast complete information) policies.
In the future, it probably also makes sense to support graphing comparisons between base learners & between protocols.
After this is done I can get to heterogenous bandit & best arm identification & work on a more interesting algorithm.