Commits

Branch selector

 master

User selector

All users

All time

Commit history

Commits on Feb 9, 2015

• 1)added normalization to the convnet version 2)fixed a bug of dimensions in nnet.train() 3) fixed a float64/32 bug in preprocessing

  neurocsut-gpu

  committed
  6853939

  Copy full SHA for 6853939

Commits on Feb 5, 2015

• preporcessor even more eficcient

  neurocsut-gpu

  committed
  6f1b495

  Copy full SHA for 6f1b495

• preprocessor 4 times more effective

  neurocsut-gpu

  committed
  7f8fd6d

  Copy full SHA for 7f8fd6d

Commits on Feb 4, 2015

• Changed approximator to 'cuda_conv' in Hybrid

  kuz

  committed
  f2a91c0

  Copy full SHA for f2a91c0

• preprocessing allows to choose between 3 different ways of cropping/resizing

  neurocsut-gpu

  committed
  0758863

  Copy full SHA for 0758863

Commits on Feb 2, 2015

• preporcessor is updated to copy deep_q_l

  neurocsut-gpu

  committed
  cd8d377

  Copy full SHA for cd8d377

• Hybrid version that allows to swich between thano and convnet in one line comment

  neurocsut-gpu

  committed
  04d8074

  Copy full SHA for 04d8074

Commits on Jan 21, 2015

• problem with hybrid not starting games solved

  neurocsut-gpu

  committed
  9ee4bcc

  Copy full SHA for 9ee4bcc

• Previously ignored shift.so file

  kuz

  committed
  e4d3a65

  Copy full SHA for e4d3a65

• Hybrid readme

  kuz

  committed
  1411507

  Copy full SHA for 1411507

• Hybrid system of our agent and Nathan Sprague's CNN, will use it to debug

  kuz

  committed
  ba156c2

  Copy full SHA for ba156c2

Commits on Dec 22, 2014

• fixed adding frames to current state

  neurocsut-gpu

  committed
  300dd17

  Copy full SHA for 300dd17

• learningrate was badly fixed

  neurocsut-gpu

  committed
  3f58bf9

  Copy full SHA for 3f58bf9

• Removed memory from ale, fixed testing issue

  neurocsut-gpu

  committed
  4c07a4b

  Copy full SHA for 4c07a4b

Commits on Dec 17, 2014

• Merge branch 'master' of ssh://github.com/kristjankorjus/Replicating-DeepMind

  kuz

  committed
  9313e21

  Copy full SHA for 9313e21

• Useful comments :)

  kuz

  committed
  b2b9413

  Copy full SHA for b2b9413

• Removed deafult parameters from ALE initialization

  kuz

  committed
  1ab0d1d

  Copy full SHA for 1ab0d1d

• Added Seaquest ROM

  kuz

  committed
  dc835a6

  Copy full SHA for dc835a6

• USE_SDL := 1 (visualisation should be on)

  kristjankorjus

  committed
  8e732e5

  Copy full SHA for 8e732e5

• Redundant definition of parameters

  kuz

  committed
  212bbb4

  Copy full SHA for 212bbb4

Commits on Dec 16, 2014

• Create folder in install script

  neurocsut-gpu

  committed
  8aabf26

  Copy full SHA for 8aabf26

• Run on first GPU device

  neurocsut-gpu

  committed
  7fcb62b

  Copy full SHA for 7fcb62b

Commits on Dec 9, 2014

• added random action for player B (inexistant), seems to make ALE more random

  RDTm

  committed
  3ed269d

  Copy full SHA for 3ed269d

Commits on Dec 8, 2014

• Set memory size so that we have enough memory on the server

  neurocsut-gpu

  committed
  70a4c96

  Copy full SHA for 70a4c96

• 6 actions instead of 4

  neurocsut-gpu

  committed
  a3cdc8c

  Copy full SHA for a3cdc8c

Commits on Dec 3, 2014

• Merge branch 'master' of github.com:kristjankorjus/Replicating-DeepMind

  tambetm

  committed
  4983be5

  Copy full SHA for 4983be5

• Fixes so that you can test without training.

  tambetm

  committed
  b8a9884

  Copy full SHA for b8a9884

• Skeleton for unit testing

  neurocsut-gpu

  committed
  45b8bbc

  Copy full SHA for 45b8bbc

• Deleted old stuff

  neurocsut-gpu

  committed
  2d04c80

  Copy full SHA for 2d04c80

• Added dummy method to SimpleDataProvider to make loading of state possible

  tambetm

  committed
  731628d

  Copy full SHA for 731628d

• Merge branch 'master' of github.com:kristjankorjus/Replicating-DeepMind

  Show description for a5fa02d
  tambetm

  committed
  a5fa02d

  Copy full SHA for a5fa02d

• Added saving of network.

  tambetm

  committed
  7b031ba

  Copy full SHA for 7b031ba

• Small changes, not tested

  neurocsut-gpu

  committed
  a335326

  Copy full SHA for a335326

• Added saving of weights.

  tambetm

  committed
  aa66bfa

  Copy full SHA for aa66bfa

• Added logging of training statistics and scores.

  Show description for 570127d
  tambetm

  committed
  570127d

  Copy full SHA for 570127d