Aiming to thouroughly benchmark and compare ML algorithms (with current focus on HTM), be designing specialized synthetic datasets that stress a single feature and can be well evaluated and understood.
For users being able to decide where each algorithm has its strong/weak-spots and decide in application for real-world problems.
This can also work as a benchmark to evaluate development impact in changes to the algorithms.
This repository should be a collection of
The topics of research interest are classified in Issue's labels, and are related to: NuPIC, dataset creation, novel research ideas, cognitive modeling, and so on...
data/datasets/*/results/
folders.datasets/generatingScripts/MAIN.m
, you'll need Matlab/R for that.python opf/anomaly_benchmark.py
NuPIC has to be installed.plotResutls/plotDatasets.m
or interactive tool nupic.visualizations, which can be run onlineresults/
subfolders under respective paths and presented in form of CSV and image data. AdaptiveScalar
encoder is NOT suitable for streaming data, use RDSE
instead. optimal resulution
for RDSEBoosting
implementation is causing artificial disturbances in the predictions (and can be improved, or should be turned off)point
anomalyinterval
anomalyWarning: anything here may, or may not be true. It is under evaluation. We are raising the topics here to get your focus on the current issues and possible findings.
AnomalyLikelihood
implementation is inferior to the "raw" Anomaly
, esp. with combination of noisy data that distort the internal distribution model. swarming
) on NuPIC's performancetrend
data?Hierarchical Temporal Memory, Numenta. Available at: http://numenta.org/resources/HTM_CorticalLearningAlgorithms.pdf
Hawkins, Jeff (2004). On Intelligence, Times Books. ISBN 0805074562.
Uhl, Christian (1999). Analysis of Neurophysiological Brain Functioning, Springer. ISBN 978-3-642-64219-7
The Sicence of Anomaly Detection, Numenta. Available at: http://numenta.com/assets/pdf/whitepapers/Numenta%20White%20Paper%20-%20Science%20of%20Anomaly%20Detection.pdf
Schmidhuber, Jürgen (2014). Deep Learning in Neural Networks: An Overview, The Swiss AI Lab IDSIA. Available at: http://arxiv.org/pdf/1404.7828v4.pdf
Twitter, Anomaly Detection. Available at: https://blog.twitter.com/2015/introducing-practical-and-robust-anomaly-detection-in-a-time-series
Skyline, Anomaly Detection. Available at: https://github.com/etsy/skyline
Yahoo, Time Series Anomaly Detection. Available at: http://yahoolabs.tumblr.com/post/114590420346/a-benchmark-dataset-for-time-series-anomaly