Skip to content

Caffe deep neural networks

Toby Dylan Hocking edited this page Dec 23, 2015 · 6 revisions

Background

Deep neural networks are powerful supervised learners for large labeled data sets. One implementation of these models is provided by the Caffe open-source C++ library.

Related work

The nnet package implements single layer neural networks. Caffe implements multi-layer (deep) neural networks. MENTORS: you should provide a detailed description of what Caffe provides over existing R packages for deep learning such as deepnet.

Other projects that use the Caffe C++ code are listed on their wiki.

Coding project: library(caffe)

Write an R package that interfaces the Caffe C++ code. It should implement the same functionality as the Python and MATLAB packages discussed on the Caffe Interfaces page.

Mentors

There are currently no mentors for this project. Ideally there should be one mentor who is an expert in R package development and another mentor who is an expert user or developer of the Caffe C++ code. Any interested students should try to find two mentors by sending emails to the caffe-users and gsoc-r lists.

Tests

Do one or several — doing more hard tests makes you more likely to be selected.

  • MENTORS: post tests for students here.

Solutions of tests

Students, please post a link to your test results here.

Clone this wiki locally