# gumby

Bitbucket is a code hosting site with unlimited public and private repositories. We're also free for small teams!

Close

# Dr. Gumby - Brain Specialist

Author: Olivier Grisel Overview of the Dr. Gumby project

## 1   Goal

Dr. Gumby is an experimental python library and set of utilities to loosely simulate neo-cortical learning and inference with a layered structure of Support Vector Machines models.

The general idea of Dr. Gumby is mainly inspired by the architectural considerations of Jeff Hawkins in his book "On Intelligence" and by the publications of Karl Friston on free energy and the brain.

The main idea is that predictions of the upper layers are used by the lowest layers to predict their future states while surprises (i.e. prediction errors) flows up from the lowest layers (next to the receptors) to the upper layers.

Experiments include 3D movie prediction and other arbitrary time series forecasting.

Dr. Gumby is implemented in python using the numpy / scipy libraries for numeric computations and plotting with a libsvm python binding to perform model training and prediction.

## 2   Getting started

Follow the instructions in INSTALL.txt to get the gumby library up and working and read the restructured text version of the documentation in the doc/ folder.

To build the HTML version using sphinx along with generated images and movies from the doctests please type:

$make test$ make html


And point your browser to doc/_build/html/index.html.

## 3   Licensing

Dr. Gumby is an Open Source project released under the GPL license.

# Recent activity

Commits by Olivier Grisel were pushed to ogrisel/gumby

13f5c9a - experimental branch to use more numpy arrays instead of python list to spare memory and calls to malloc
262f72c - fix style + train a model by default
91984b1 - smaller default dimensions + added timing information
9b72a63 - ignore ctags index file
d6a3810 - use 32bit floats by default to spare some memory
898a220 - make it possible to store the training set of a 1 layer experiment for later usage
f5d27b8 - make Layer.build_training_set() public
354a8e6 - make it possible to use another directory for test movies in xp_movie_prediction.py
a85ac32 - factorise out the method to build training sets out of brain history
Tip: Filter by directory path e.g. /media app.js to search for public/media/app.js.
Tip: Use camelCasing e.g. ProjME to search for ProjectModifiedEvent.java.
Tip: Filter by extension type e.g. /repo .js to search for all .js files in the /repo directory.
Tip: Separate your search with spaces e.g. /ssh pom.xml to search for src/ssh/pom.xml.
Tip: Use ↑ and ↓ arrow keys to navigate and return to view the file.
Tip: You can also navigate files with Ctrl+j (next) and Ctrl+k (previous) and view the file with Ctrl+o.
Tip: You can also navigate files with Alt+j (next) and Alt+k (previous) and view the file with Alt+o.