Commits

Branch selector

 master

User selector

efiko

All time

Commit History

Commits on Oct 31, 2022

• Explicitly cast the Boltzmann Gumbel exploration coefficient to float before using it.

  

  efiko

  authored and

  Copybara-Service

  committedOct 31, 2022
  a1cd44d

• Small fix in the train_eval_stationary_linear script

  

  efiko

  authored and

  Copybara-Service

  committedOct 31, 2022
  684f910

Commits on Aug 26, 2022

• Add support for Laplacian smoothing in the multi-objective agent.

  

  efiko

  authored and

  Copybara-Service

  committedAug 26, 2022
  2cbf713

Commits on Jun 3, 2022

• Add support for having an activation in the Q layer.

  

  efiko

  authored and

  Copybara-Service

  committedJun 3, 2022
  4b2203f

Commits on Apr 13, 2022

• Add support to override the observation spec in trainable constraints.

  

  efiko

  authored and

  Copybara-Service

  committedApr 13, 2022
  13299f8

Commits on Apr 1, 2022

• Add support for overriding the network in trainable constraints.

  

  efiko

  authored and

  Copybara-Service

  committedApr 1, 2022
  5909388

Commits on Nov 24, 2021

• Add support to disable the policy saver in the trainer.

  

  efiko

  authored and

  Copybara-Service

  committedNov 24, 2021
  11d3b98

Commits on Nov 23, 2021

• Fixes the policy_info of the Bernoulli Thompson Sampling policy.

  

  efiko

  authored and

  Copybara-Service

  committedNov 23, 2021
  da4aa03

Commits on Nov 22, 2021

• Train & eval binary on Bernoulli bandit environment.

  

  efiko

  authored and

  Copybara-Service

  committedNov 22, 2021
  fcc5013

Commits on Nov 16, 2021

• Bernoulli Thompson Sampling agent

  

  efiko

  authored and

  Copybara-Service

  committedNov 16, 2021
  8042e71

Commits on Nov 15, 2021

• Bernoulli Thompson sampling policy

  

  efiko

  authored and

  Copybara-Service

  committedNov 15, 2021
  f721bb5

Commits on Nov 2, 2021

• Switch to using tf.int64 for holding the number of samples per action.

  

  efiko

  authored and

  Copybara-Service

  committedNov 2, 2021
  c060bf0

• Make the Boltzmann-Gumbel agent XLA-compatible.

  

  efiko

  authored and

  Copybara-Service

  committedNov 2, 2021
  f2fbd29

• Add the Boltzmann-Gumbel agent to the stationary linear environment tests.

  

  efiko

  authored and

  Copybara-Service

  committedNov 2, 2021
  db66f82

Commits on Oct 29, 2021

• Neural Bandit with Boltzmann-Gumbel exploration.

  

  efiko

  authored and

  Copybara-Service

  committedOct 29, 2021
  cbc5c16

Commits on Jun 1, 2021

• Example usage of the Neural Boltzmann agent.

  

  efiko

  authored and

  Copybara-Service

  committedJun 1, 2021
  f13d3a9

• Neural bandit agent with Boltzmann exploration

  

  efiko

  authored and

  Copybara-Service

  committedJun 1, 2021
  b8c2981

Commits on May 31, 2021

• Bandit reward prediction policy with Boltzmann exploration

  

  efiko

  authored and

  Copybara-Service

  committedMay 31, 2021
  9729df6

Commits on Nov 2, 2020

• Add an InputNetworkConstraint that represents a constraint using a (trained) input network.

  

  efiko

  authored and

  Copybara-Service

  committedNov 2, 2020
  faaedcb

Commits on Oct 27, 2020

• Fix a corner-case bug in epsilon greedy policy, occurring when epsilon is 0.

  

  efiko

  authored and

  Copybara-Service

  committedOct 27, 2020
  4950a6d

Commits on Sep 3, 2020

• Fix broken link in the bandits README file.

  

  efiko

  authored and

  Copybara-Service

  committedSep 3, 2020
  32cca86

• Add a README file in the bandits directory

  

  efiko

  authored and

  Copybara-Service

  committedSep 3, 2020
  6d4d2df

Commits on Aug 28, 2020

• Add type annotations to the Bandit environments.

  

  efiko

  authored and

  Copybara-Service

  committedAug 28, 2020
  d68902e

Commits on Jul 16, 2020

• Make the restart() function more robust for non-scalar rewards.

  

  efiko

  authored and

  Copybara-Service

  committedJul 16, 2020
  c5a2f1f

Commits on Jun 26, 2020

• Add a simple absolute constraint.

  

  efiko

  authored and

  Copybara-Service

  committedJun 26, 2020
  1a39da4

Commits on Jun 23, 2020

• Add a new relative constraint acting on expected values.

  

  efiko

  authored and

  Copybara-Service

  committedJun 23, 2020
  737b8cc

Commits on Jun 12, 2020

• Make the reward_spec nicely structured when we have constraints.

  

  efiko

  authored and

  Copybara-Service

  committedJun 12, 2020
  78c163f

Commits on Jun 9, 2020

• Add constraint violations as a new metric in bandit tf-metrics.

  

  efiko

  authored and

  Copybara-Service

  committedJun 9, 2020
  751cafa

Commits on Jun 8, 2020

• Introduce a constraint class for expressing relative quantile constraints with respect to a baseline action/policy.

  

  efiko

  authored and

  Copybara-Service

  committedJun 8, 2020
  b2d0258

Commits on Jun 4, 2020

• Extend the regret metric to work with nested rewards.

  

  efiko

  authored and

  Copybara-Service

  committedJun 4, 2020
  23727e3

• Fix the SparseObs test in NeuralLinUCB policy.

  

  efiko

  authored and

  Copybara-Service

  committedJun 4, 2020
  d1caea1

• Add support for distributed training of the NeuralLinUCB agent.

  

  efiko

  authored and

  Copybara-Service

  committedJun 4, 2020
  8bc5d95

Commits on May 28, 2020

• Update the TFPyEnvironment to support rewards with a nested structure.

  

  efiko

  authored and

  Copybara-Service

  committedMay 28, 2020
  a39f922

Commits on May 27, 2020

• Update the StationaryStochasticPyEnvironment to produce structured rewards that are composed of 'reward' and 'constraint' parts.

  

  efiko

  authored and

  Copybara-Service

  committedMay 27, 2020
  2e616ba

Commits on May 19, 2020

• Add support for constrained optimization in Neural Bandits agents.

  

  efiko

  authored and

  Copybara-Service

  committedMay 19, 2020
  6eae214

Pagination

PreviousNext