Tokyo Machine Learning Society

Organizing : Reactive Inc.

Group Description

TMLS is focused on developing machine intelligence in Japan and specifically focused on deep learning developments and challenges.

Three core factors have enabled recent advances in deep learning providing for its ‘unusual efficiency’. These factors are:

the explosion in large data sets along with the knowledge that supervised learning this data could achieve test set generalization

high performance computation architectures, build on GPU grid architectures and especially CUDA and Nvidia technologies have allowed for faster, cheaper learning with these very large data sets

and the development of new powerful algorithms such as rectified linear hidden units or new regularisation techniques (Dropout), but also new topologies provide the framework to learn efficiently

TMLS will look at applications into numerous spaces and we are particularly interested in visual and speech recognition, reinforcement learning (RL + DL = AI?), natural language processing, medical and health applications and financial engineering.

We will take a look at all the leading technologies from Spark and H20 (CPU learning) to Distributed learning systems like Neon, Torch and Caffe.

TMLS is sponsored by Reactive Inc. (Tokyo).

Please contact us if you would like to speak or sponsor an event.

Finished Events View all events (4)

Ended 2016/07/29(Fri) 19:00〜

TMLS#4 Review of the state of the art

Chie Matsunaga Chie Matsunaga


Up to 120

Ended 2016/05/27(Fri) 19:00〜

TMLS #3 Applying Deep Learning

Chie Matsunaga Chie Matsunaga


Up to 100

Ended 2016/03/25(Fri) 18:30〜


Chie Matsunaga Chie Matsunaga



Members 278


  • Chie Matsunaga
  • nishimurakaz
  • Reiko Hata

Other Members

  • y
  • Yooskee_a
  • Yoshi
  • rheza_h
  • GitHub30
  • kaeken
  • HiromuMasuda
  • keiyoo
  • Tetsuo Seto
  • Tomato_Potato
  • d_maeno
  • wastedlives
  • YurenZhang
  • NaoyukiTomita
  • Hito
  • harupy
  • knife0125
  • arvindpal
  • 坂本航太郎
  • ChrisNeuman



    If you add media to group events the most recent 10 media items are shown.