cosmicBboy/ml-research

Issues

• metalearn swarm implementation

  #27 opened 4 years ago by cosmicBboy
  0

• [metalearn] neurips bbo challenge idea dump

  #26 opened 4 years ago by cosmicBboy
  4

• [metalearn] support PPO training algorithm

  #25 opened 4 years ago by cosmicBboy
  1

• [metalearn] support arbitrary expansion of action space in metalearn_controller

  #24 opened 4 years ago by cosmicBboy
  1

• [metalearn] support continuous policy action space

  #23 opened 4 years ago by cosmicBboy
  1

• [metalearn] make sure log action probs before masking is correct

  #16 opened 5 years ago by cosmicBboy
  0

• [metalearn] RNN hidden state should not be re-initialized on each fit_iter call

  #17 opened 5 years ago by cosmicBboy
  0

• [metalearn]: catch RuntimeError where all action probabilities are 0

  #15 opened 5 years ago by cosmicBboy
  0

• [metalearn] correctly handle scorers where needs_proba=True

  #11 opened 5 years ago by cosmicBboy
  0

• [metalearn] implement advantage actor critic algorithm

  #19 opened 5 years ago by cosmicBboy
  1

• [metalearn] run experiment with fewer number of tasks

  #22 opened 5 years ago by cosmicBboy
  0

• [metalearn] look into coefficient of value function loss

  #21 opened 5 years ago by cosmicBboy
  0

• [metalearn] look into adding current timestep as input

  #20 opened 5 years ago by cosmicBboy
  0

• use -inf for exclusion mask implementation, make sure algorithm actions cannot be masked/ignored

  #18 opened 5 years ago by cosmicBboy
  1

• [metalearn] create daemon-compatible pipeline fit limit enforcer

  #14 opened 5 years ago by cosmicBboy
  3

• [metalearn] explore use of Prince package for factor analysis

  #13 opened 5 years ago by cosmicBboy
  0

• [metalearn] columntransformer re-arranges the input matrix based on column indices specified in transformers arg

  #12 opened 5 years ago by cosmicBboy
  1

• [metalearn] implement raw metafeature encoder

  #10 opened 5 years ago by cosmicBboy
  0

• [metalearn] add more feature to task state

  #9 opened 5 years ago by cosmicBboy
  0

• [metalearn] test for meta-learning capabilities: task-type-specific environments

  #8 opened 5 years ago by cosmicBboy
  0

• [metalearn] 100% coverage of sklearn API

  #6 opened 5 years ago by cosmicBboy
  0

• [metalearn] add support for XGboost, LiteGBM, CatBoost

  #7 opened 5 years ago by cosmicBboy
  0

• metalearn: ensembling controller

  #5 opened 5 years ago by cosmicBboy
  0

• add apricot module for dataset subsampling

  #3 opened 6 years ago by cosmicBboy
  0

• add support for normalizing the reward signal before computing gradients

  #2 opened 6 years ago by cosmicBboy
  1

• Bug: baseline_reward_buffer updating logic leads to incorrect implementation

  #1 opened 6 years ago by cosmicBboy
  0