Open Source DeepMind
On 18 May, 2015 By michaelmalak 0 Comments
Well, not exactly DeepMind -- for surely their technology has progressed -- but there are now several open source implementations of the algorithms described in Neural Turing Machines by Graves et al. The above diagram comes from an open source implementation described at awawfumin.blogspot.com.
Recall that a Neural Turning Machine (popularized by the company, DeepMind, which Google acquired) is not simply deep learning (deep neural network), but is a deep neural network that also includes memory storage. Recall that a DeepMind system has been demonstrated to learn to play 1980's videogames.
There are several implementations available for download on GitHub, so that you can have your very own artificial "mind":
- The implementation in Go described above.
- An implementation in Python by Shawn Tan at the National University of Singapore that the author of the Go implementation referred to.
- An implementation in Torch/Lua by Kai Sheng Tai at Stanford
- An implementation in Node.js by Casey Gibson at IBM