Beginner's Guide to NuPIC
Download as PPTX, PDF3 likes5,426 views
This document provides an overview of NuPIC, an open source machine learning framework. It discusses the different levels of abstraction in NuPIC including algorithms, networks and regions, and the online prediction framework. It also describes how NuPIC supports arbitrary network topologies through its network and regions API and handles data flow through different regions in a network.
1 of 5
Downloaded 94 times
Beginner's Guide to NuPIC
- 1. Beginner’s Guide to Developing with NuPIC Scott Purdy May 3, 2014
- 2. Levels of Abstraction • Algorithms – Encoders – Spatial Pooler – Sequence Memory (Temporal Pooler) – CLA Classifier (not covered) • Networks and Regions – supports topology • Online Prediction Framework (OPF) – CLAModel (network and algorithm wrapper) – OpfRunExperiment.py – run_swarm.py – experiment.json (not covered)
- 3. Interactive work with ipython notebook: http://nbviewer.ipython.org/url/fer.io/~scott/nu pic_overview.ipynb
- 4. Network and Regions API • Supports arbitrary topology • Formalizes serialization (but delegates implementation to each region) • Handles data flow through network
- 5. Network and Regions API Classifier L2 Temporal Pooler L2 Spatial Pooler L1 Temporal Pooler L1 Spatial Pooler Audio L1 Temporal Pooler L1 Spatial Pooler Image