megaman: Manifold Learning for Millions of Points
megaman
is a scalable manifold learning package implemented in
python. It has a front-end API designed to be familiar
to scikit-learn but harnesses
the C++ Fast Library for Approximate Nearest Neighbors (FLANN)
and the Sparse Symmetric Positive Definite (SSPD) solver
Locally Optimal Block Precodition Gradient (LOBPCG) method
to scale manifold learning algorithms to large data sets.
On a personal computer megaman can embed 1 million data points
with hundreds of dimensions in 10 minutes.
megaman is designed for researchers and as such caches intermediary
steps and indices to allow for fast re-computation with new parameters.
Package documentation can be found at http://mmp2.github.io/megaman/
If you use our software please cite the following JMLR paper:
McQueen, Meila, VanderPlas, & Zhang, "Megaman: Scalable Manifold Learning in Python", Journal of Machine Learning Research, Vol 17 no. 14, 2016. http://jmlr.org/papers/v17/16-109.html
You can also find our arXiv paper at http://arxiv.org/abs/1603.02763
Examples
Installation and Examples in Google Colab
Below it's a tutorial to install megaman on Google Colab, through Conda environment.
It also provides tutorial of using megaman to build spectral embedding on uniform swiss roll dataset.
Installation with Conda
Due to the change of API,
$ conda install -c conda-forge megaman
is no longer supported.
We are currently working on fixing the bug.
Please see the full install instructions below to build megaman
from source.
Installation from source
To install megaman from source requires the following:
- python tested with versions 2.7, 3.5 and 3.6
- numpy version 1.8 or higher
- scipy version 0.16.0 or higher
- scikit-learn
- FLANN
- pyflann which offers another method of computing distance matrices (this is bundled with the FLANN source code)
- cython
- a C++ compiler such as
gcc
/g++
Optional requirements include
- pyamg, which allows for faster decompositions of large matrices
- nose for running the unit tests
- h5py for reading testing .mat files
- plotly an graphing library for interactive plot
These requirements can be installed on Linux and MacOSX using the following conda command:
$ conda create -n manifold_env python=3.5 -y
# can also use python=2.7 or python=3.6
$ source activate manifold_env
$ conda install --channel=conda-forge -y pip nose coverage cython numpy scipy \
scikit-learn pyflann pyamg h5py plotly
Clone this repository and cd
into source repository
$ cd /tmp/
$ git clone https://github.com/mmp2/megaman.git
$ cd megaman
Finally, within the source repository, run this command to install the megaman
package itself:
$ python setup.py install
Unit Tests
megaman uses nose
for unit tests. With nose
installed, type
$ make test
to run the unit tests. megaman
is tested on Python versions 2.7, 3.4, and 3.5.
Authors
Other Contributors
- Xiao Wang: lazy rmetric, Nystrom Extension
- Hangliang Ren (Harry): Installation tutorials, Spectral Embedding
Future Work
See this issues list for what we have planned for upcoming releases: