Skip to main content

Tree-based machine learning for embedded system

Project description

emtrees

Tree-based machine learning classifiers for microcontroller and embedded systems. Train in Python, then do inference on any device with support for C.

Want Naive Bayes instead? Go to embayes

Key features

Embedded-friendly Inference

  • Portable C99 code
  • No libc required
  • No dynamic allocations
  • Integer/fixed-point math only
  • Single header file include
  • Fast, sub-millisecond classification
  • Memory efficient. Can run with <100 bytes RAM

Convenient Training

  • API-compatible with scikit-learn
  • Implemented in Python 3
  • C classifier accessible in Python using pybind11

MIT licensed

Can be used as an open source alternative to MATLAB Classification Trees, Decision Trees using MATLAB Coder for C/C++ code generation. fitctree, fitcensemble, TreeBagger, ClassificationEnsemble, CompactTreeBagger

Status

Minimally useful

  • Random Forests and ExtraTrees classifiers implemented
  • Tested running on AVR Atmega, ESP8266 and Linux.
  • On ESP8266, 8x8 digits classify in under 0.3ms with 95%+ accuracy
  • On Linux, is approx 2x faster than sklearn

Installing

Install from PyPI

pip install emtrees --user

Usage

  1. Train your model in Python
import emtrees
estimator = emtrees.RandomForest(n_estimators=10, max_depth=10)
estimator.fit(X_train, Y_train)
...
  1. Generate C code
code = estimator.output_c('sonar')
with open('sonar.h', 'w') as f:
   f.write(code)
  1. Use the C code
#include <emtrees.h>
#include "sonar.h"

const int32_t length = 60;
int32_t values[length] = { ... };
const int32_t predicted_class = sonar_predict(values, length):

For full example code, see examples/digits.py and emtrees.ino

TODO

0.2

  • Standalone example application on microcontroller
  • Include emtrees.h inline in generated code

1.0

  • Support returning probabilities
  • Support serializing/deserializing trees

Probably

  • Support sklearn GradientBoostingClassifier
  • Support regression trees
  • Support weighted voting
  • Implement Isolation Forests (requires path/depths)

Maybe

  • Support XGBoost learning of trees
  • Support LightGBM learning of trees
  • Support CatBoost learning of trees
  • Support/Implement a Very Fast Decision Tree (VFDT) learning algorithm
  • Implement multithreading when used in Python bindings, using OpenMP

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

emtrees-0.2.5.tar.gz (13.6 kB view details)

Uploaded Source

File details

Details for the file emtrees-0.2.5.tar.gz.

File metadata

  • Download URL: emtrees-0.2.5.tar.gz
  • Upload date:
  • Size: 13.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/1.12.1 pkginfo/1.4.2 requests/2.20.0 setuptools/40.4.3 requests-toolbelt/0.8.0 tqdm/4.27.0 CPython/3.6.3

File hashes

Hashes for emtrees-0.2.5.tar.gz
Algorithm Hash digest
SHA256 037e7a9b7b6f21aefb8e5e8d00863414d4784d4a9755d6278dab2390e3c7e916
MD5 101e20598a1cc4c1bc03188ae7d93e22
BLAKE2b-256 6417f1fcf11fde45023e787d271df0a1c5c890a887e7dcdabd3c646b40765ead

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page