Edge Machine Learning
This repository provides code for machine learning algorithms for edge devices developed at Microsoft Research India.
Machine learning models for edge devices need to have a small footprint in terms of storage, prediction latency and energy. One example of a ubiquitous real-world application where such models are desirable is resource-scarce devices and sensors in the Internet of Things (IoT) setting. Making real-time predictions locally on IoT devices without connecting to the cloud requires models that fit in a few kilobytes.
This repository contains two such algorithms Bonsai and ProtoNN that shine in this setting. These algorithms can train models for classical supervised learning problems with memory requirements that are orders of magnitude lower than other modern ML algorithms. The trained models can be loaded onto edge devices such as IoT devices/sensors, and used to make fast and accurate predictions completely offline.
We welcome contributions, comments and criticism. For questions, please email Harsha.
- gcc version 5.4. Other gcc versions above 5.0 could also work.
- We developed the code on Ubuntu 16.04LTS. Other linux versions could also work.
- You can either use the Makefile in the root, or cmake via the build directory (see below).
- Visual Studio 2015. Use cmake (see below).
- For Anniversary Update or later, one can use the Windows Subsystem for Linux, and the instructions for Linux build.
On both Linux and Windows 10, you need an implementation of BLAS, sparseBLAS and vector math calls. We link with the implementation provided by the Intel(R) Math Kernel Library. Please download later versions (2017v3+) of MKL as far as possible. The code can be made to work with other math libraries with a few modifications.
Building using Makefile
After cloning this repository, set compiler and flags appropriately in
config.mk. Then execute the following in bash:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<MKL_PATH>:<EDGEML_ROOT> make -Bj
Typically, MKL_PATH = /opt/intel/mkl/lib/intel64_lin/, and EDGEML_ROOT is '.'.
This will build four executables BonsaiTrain, BonsaiPredict, ProtoNNTrain and ProtoNNPredict in <EDGEML_ROOT>. Sample data to try these executables is not included in this repository, but instructions to do so are given below.
Building using CMake
For Linux, in the <EDGEML_ROOT> directory:
mkdir build export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<MKL_PATH> cd build cmake .. make -Bj
For Windows 10, in the <EDGEML_ROOT> directory, modify
CMakeLists.txt file to change <MKL_ROOT> by changing the
Then, generate Visual Studio 2015 solution using:
cd build cmake -G "Visual Studio 14 2015 Win64" -DCMAKE_BUILD_TYPE=Release ..
EdgeML.sln in VS2015, build and run.
For both Linux and Windows10, cmake builds will generate four executables BonsaiTrain, BonsaiPredict, ProtoNNTrain and ProtoNNPredict in <EDGEML_ROOT>.
Download a sample dataset
Follow the bash commands given below to download a sample dataset, USPS10, to the root of the repository. Bonsai and ProtoNN come with sample scripts to run on the usps10 dataset. EDGEML_ROOT is defined in the previous section.
cd <EDGEML_ROOT> mkdir usps10 cd usps10 wget http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/usps.bz2 wget http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/usps.t.bz2 bzip2 -d usps.bz2 bzip2 -d usps.t.bz2 mv usps train.txt mv usps.t test.txt mkdir ProtoNNResults cd <EDGEML_ROOT>
This will create a sample train and test dataset, on which you can train and test Bonsai and ProtoNN algorithms. As specified, we create an output folder for ProtoNN. Bonsai on the other hand creates its own output folder. For instructions to actually run the algorithms, see Bonsai Readme and ProtoNN Readme.
You could change the behavior of the code by setting these flags in
config.mk and rebuilding with
make -Bj when building with the default Makefile in <EDGEML_ROOT>. When building with CMake, change these flags in
CMakeLists.txt in <EDGEML_ROOT>. All these flags can be set for both ProtoNN and Bonsai.
The following are supported currently by both ProtoNN and Bonsai.
SINGLE/DOUBLE: Single/Double precision floating-point. Single is most often sufficient. Double might help with reproducibility. ZERO_BASED_IO: Read datasets with 0-based labels and indices instead of the default 1-based. TIMER: Timer logs. Print running time of various calls. CONCISE: To be used with TIMER to limit the information printed to those deltas above a threshold.
The following currently only change the behavior of ProtoNN, but one can write corresponding code for Bonsai.
LOGGER: Debugging logs. Currently prints min, max and norm of matrices. LIGHT_LOGGER: Less verbose version of LOGGER. Can be used to track call flow. XML: Enable training with large sparse datasets with many labels. This is in beta. VERBOSE: Print additional informative output to stdout. DUMP: Dump models after each optimization iteration instead of just in the end. VERIFY: Legacy verification code for comparison with Matlab version.
Additionally, there is one of two flags that has to be set in the Makefile:
MKL_PAR_LDFLAGS: Linking with parallel version of MKL. MKL_SEQ_LDFLAGS: Linking with sequential version of MKL.