yandexdataschool/AgentNet

Issues

• Original DQN Example

  #103 opened 7 years ago by ehknight
  1

• Destination GpuArray is not contiguous

  #96 opened 7 years ago by kashif
  5

• attention tests

  #99 opened 7 years ago by justheuristic
  1

• policy_estimators param is weird

  #93 opened 7 years ago by pshvechikov
  3

• example: Learning to be kind

  #48 opened 7 years ago by justheuristic
  1

• Targetnet of layer on top of LSTMCell results in deepcopy error

  #94 opened 7 years ago by pshvechikov
  3

• deprecate preprocess_observation

  #90 opened 7 years ago by justheuristic
  1

• batch_size parameter is wierd

  #95 opened 7 years ago by pshvechikov
  1

• Support both Theano (Lasagne; Keras) and Tensorflow (Keras) backend

  #92 opened 7 years ago by Omrigan
  1

• BaseResolver returns int64

  #91 opened 7 years ago by justheuristic
  1

• Vectorized environment

  #89 opened 7 years ago by justheuristic
  0

• DPG refactor and demo

  #86 opened 7 years ago by justheuristic
  1

• example:Qlearning with normalized advantage functions

  #81 opened 7 years ago by justheuristic
  1

• grad dtypes mismatch in some rare case

  #83 opened 7 years ago by justheuristic
  2

• Deprecation list

  #68 opened 7 years ago by justheuristic
  0

• better weights management for memory layers

  #84 opened 8 years ago by justheuristic
  0

• AgentNet recurrence won't compile if batch_size = 1 and unroll_scan=False and at least one input is a single-element vector.

  #79 opened 8 years ago by justheuristic
  2

• canonicalize LSTM

  #80 opened 8 years ago by justheuristic
  3

• Brief outline of modules

  #77 opened 8 years ago by arogozhnikov
  3

• Hierarchical MDP as a demo?

  #71 opened 8 years ago by justheuristic
  0

• Counter, switch, everyK (see if it works and saves time)

  #49 opened 8 years ago by justheuristic
  0

• Minimal initial example.

  #53 opened 8 years ago by arogozhnikov
  5

• Automated tests on convergence

  #54 opened 8 years ago by arogozhnikov
  3

• KSfinder experiment setup

  #32 opened 8 years ago by justheuristic
  1

• Dockerfile aka "makeitwork"

  #69 opened 8 years ago by justheuristic
  0

• TupleLayer refactor

  #55 opened 8 years ago by justheuristic
  1

• TODOs

  #58 opened 8 years ago by justheuristic
  0

• Continuous action space policy gradient

  #50 opened 8 years ago by justheuristic
  1

• Add py3 to container

  #52 opened 8 years ago by justheuristic
  0

• Continuous/ndimensional action support

  #25 opened 8 years ago by justheuristic
  2

• Recurrence internal refactor as Lasagne layer

  #47 opened 8 years ago by justheuristic
  0

• Environment interface with Lasagne layers

  #46 opened 8 years ago by justheuristic
  1

• Environment model agent training (a.k.a learning curiosity)

  #31 opened 8 years ago by justheuristic
  1

• Adversarial architecture

  #35 opened 8 years ago by justheuristic
  1

• Release preparations

  #34 opened 8 years ago by justheuristic
  1

• Forced category predictions

  #40 opened 8 years ago by justheuristic
  1

• persistence: support unnamed layers and layers with same names

  #29 opened 8 years ago by justheuristic
  5

• Allow environments that work outside theano + make tutorial

  #37 opened 8 years ago by justheuristic
  1

• Window memory

  #44 opened 8 years ago by justheuristic
  1

• Getting published

  #33 opened 8 years ago by justheuristic
  2

• Store initial hidden values with SessionPool and SessionBatch

  #43 opened 8 years ago by justheuristic
  1

• Dialogs demo stand

  #41 opened 8 years ago by justheuristic
  1

• could not download data to installation folder

  #42 opened 8 years ago by justheuristic
  2

• Remove "Qvalues"-related names from where inappropriate

  #38 opened 8 years ago by justheuristic
  3

• A3c a.k.a. Actor-Critic method

  #36 opened 8 years ago by justheuristic
  2

• Learning refactor

  #39 opened 8 years ago by justheuristic
  1

• K-step reinforcement learning

  #27 opened 8 years ago by justheuristic
  4

• Reinforcement Learning Comparison

  #28 opened 8 years ago by justheuristic
  2

• Session printing broken

  #30 opened 8 years ago by justheuristic
  1

• Implement SARSA and compare with Q-learning

  #26 opened 8 years ago by justheuristic
  2