Skip to content
This repository has been archived by the owner on Aug 5, 2022. It is now read-only.

Latest commit

 

History

History
480 lines (309 loc) · 12.3 KB

README.md

File metadata and controls

480 lines (309 loc) · 12.3 KB
Raw

DISCONTINUATION OF PROJECT.

This project will no longer be maintained by Intel.

Intel has ceased development and contributions including, but not limited to, maintenance, bug fixes, new releases, or updates, to this project.

Intel no longer accepts patches to this project.

If you have an ongoing need to use this project, are interested in independently developing it, or would like to maintain patches for the open source software community, please create your own fork of this project. Table of Contents

Intel® Pattern Matching Technology

This repository contains the CuriePME library, which provides access to the Pattern Matching Engine (PME) within the Intel® Curie™ Compute Module.

Supported Curie™ hardware platforms:

Development Environments

This library can be used in the following development environments:

  • On Ubuntu 14.04 LTS or 16.04 LTS, using the latest version of the Curie Open Developer Kit

  • In the Arduino IDE ("Download as a ZIP", unzip, place inside your libraries folder)

About the Library

the PME is a hardware engine capable of learning and recognizing patterns in arbitrary sets of data. The CuriePME library provides access to the PME, making it possible to implement machine-learning pattern matching or classification algorithms which are accelerated by the pattern-matching capabilities of the PME.

  • Basic Functions Supported:
    • Learning Patterns
    • Recognizing Patterns
    • Storing Pattern Memory (Knowledge) [Requires SerialFlash Library]
    • Retrieving Pattern Memory (Knowledge) [Requires SerialFlash Library]

About the Intel® Curie™ Pattern Matching Engine

The Pattern Matching Engine (PME) is a parallel data recognition engine with the following features:

  • 128 parallel Processing Elements (PE) each with"

    • 128 byte input vector

    • 128 byte model memory.

    • 8-Bit Arithmetic Units

    • Two distance evaluation norms with 16-bit resolution:

      • L1 norm (Manhattan Distance)
      • Lsup (Supremum) norm (Chebyshev Distance)

    • Support for up to 32,768 Categories

    • Classification states:

       * ID  - Identified
 * UNC - Uncertain
 * UNK - Unknown

  • Two Classification Functions:

    • k-nearest neighbors (KNN)
    • Radial Basis Function (RBF)

  • Support for up to 127 Contexts

CuriePME API reference

Constants

  • CuriePME.noMatch (uint32_t): The value returned by classify() to indicate a pattern could not be classified
  • CuriePME.minContext (uint16_t): Minimum context value
  • CuriePME.maxContext (uint16_t): Maximum context value
  • CuriePME.maxVectorSize (int32_t): Maximum pattern size (in bytes) that can be accepted by learn() and classify()
  • CuriePME.firstNeuronID (int32_t): ID of first neuron in network
  • CuriePME.lastNeuronID (int32_t): ID of last neuron in network
  • CuriePME.maxNeurons (int32_t): Number of neurons in the network

Initialization Functions

CuriePME.begin()

void CuriePME.begin(void)

Initialize the PME so it is ready for operation

Parameters

none

Return value

none

CuriePME.forget()

void CuriePME.forget(void)

Clear any data committed to the network, making the network ready to learn again.

Parameters

none

Return value

none

Basic Functions

CuriePME.learn()

uint16_t CuriePME.learn(uint8_t *pattern_vector, int32_t vector_length, uint8_t category)

Takes a pattern pattern_vector of size vector_length, and commits it to the network as training data. The category parameter indicates to the PME which category this training vector belongs to- that is, if a future input has a sufficiently similar pattern, it will be classified as the same category passed with this pattern.

Parameters

  1. uint8_t *pattern_vector : Pointer to the training data. Training data must be no longer than 128 bytes
  2. int32_t vector_length : The size (in bytes) of your training vector
  3. uint8_t category : The category that should be assigned to this data

Return value

Total number of committed neurons in the network after the learning operation is complete

CuriePME.classify()

uint16_t CuriePME.classify(uint8_t *pattern_vector, int32_t vector_length)

Takes a pattern pattern_vector of size vector_length, and uses the committed neurons in the network to classify the pattern

Parameters

  1. uint8_t *pattern_vector : Pointer to the data to be classified. Pattern data must be no longer than 128 bytes
  2. int32_t vector_length : The size (in bytes) of the data to be classified

Return value

CuriePME.noMatch if the input data did not match any of the trained categories. Otherwise, the trained category assigned by the network will be returned

Saving Knowledge

CuriePME.beginSaveMode()

void CuriePME.beginSaveMode(void)

Puts the network into a state that allows the neuron contents to be read

Parameters

none

Return value

none

CuriePME.iterateNeuronsToSave()

uint16_t CuriePME.iterateNeuronsToSave(neuronData& data_array)

When in save mode, this method can be called to obtain the data for the next committed neuron. Each successive call will increment an internal pointer and return the data for successive committed neurons, until all committed neurons have been read.

Parameters

  1. neuronData& data_array : a neuronData type in which the neuron data will be placed

Return value

0 when all committed neurons have been read. Otherwise, this method returns the trained category of the neuron being read

CuriePME.endSaveMode()

void CuriePME.endSaveMode(void)

Takes the network out of save mode

Parameters

none

Return value

none

Restoring Knowledge

CuriePME.beginRestoreMode()

void CuriePME.beginRestoreMode(void)

Puts the network into a state that allows the neuron contents to be restored from a file

Parameters

none

Return value

none

CuriePME.iterateNeuronsToRestore()

void CuriePME.iterateNeuronsToRestore(neuronData& data_array)

When in restore mode, this method can be called to write data to the next available neuron. Each successive call will increment an internal pointer until all the neurons in the network have been written.

Parameters

  1. neuronData& data_array : a neuronData type containing the neuron data

Return value

none

CuriePME.endRestoreMode()

void CuriePME.endRestoreMode(void)

Takes the network out of restore mode

Parameters

none

Return value

none

Configuraton Functions

CuriePME.setClassifierMode()

void CuriePME.setClassifierMode(PATTERN_MATCHING_CLASSIFICATION_MODE mode)

Sets the classifying function to be used by the network

Parameters

  1. PATTERN_MATCHING_CLASSIFICATION_MODE mode The classifying function to use. Valid values are:
    • RBF_Mode (default)
    • KNN_Mode

Return value

none

CuriePME.getClassifierMode()

PATTERN_MATCHING_CLASSIFICATION_MODE CuriePME.getClassifierMode(void)

Gets the classifying function being used by the network

Parameters

none

Return value

PATTERN_MATCHING_CLASSIFICATION_MODE mode The classifying function being used. Possible values are:

  • RBF_Mode
  • KNN_Mode

CuriePME.setDistanceMode()

void CuriePME.setDistanceMode(PATTERN_MATCHING_DISTANCE_MODE mode)

Sets the distance function to be used by the network

Parameters

  1. PATTERN_MATCHING_DISTANCE_MODE mode The distance function to use. Valid values are:

    • LSUP_Distance (default)
    • L1_Distance

Return value

none

CuriePME.getDistanceMode()

PATTERN_MATCHING_DISTANCE_MODE CuriePME.getDistanceMode(void)

Gets the distance function being used by the network

Parameters

none

Return value

PATTERN_MATCHING_DISTANCE_MODE mode The distance function being used. Possible values are:

  • LSUP_Distance
  • L1_Distance

CuriePME.setGlobalContext()

void CuriePME.setGlobalContext(uint16_t context)

Writes a value to the Global Context Register. Valid context values range between 1-127. A context value of 0 enables all neurons, with no regard to their context

Parameters

  1. uint16_t context : a value between 0-127 representing the desired context. A context value of 0 selects all neurons, regardless of their context value.

Return value

none

CuriePME.getGlobalContext()

uint16_t CuriePME.getGlobalContext(void)

Reads the Global Context Register

Parameters

none

Return value

uint16_t, the contents of the Global Context Register (a value between 0-127)

Other Functions

CuriePME.getCommittedCount()

uint16_t CuriePME.getCommittedCount(void)

Gets the number of committed neurons in the network (NOTE: this method should not be used in save/restore mode, because it will give inaccurate results)

Parameters

none

Return value

uint16_t, the number of comitted neurons in the network

CuriePME.readNeuron()

void CuriePME.readNeuron(int32_t neuronID, neuronData& data_array)

Read a specific neuron by its ID

Parameters

  1. int32_t neuronID : value between 1-128 representing a specific neuron
  2. neuronData& data_array : neuronData type in which to write the neuron data

Return value

none

CuriePME.writeVector() (KNN_Mode only)

uint16_t CuriePME.writeVector(uint8_t *pattern_vector, int32_t vector_length)

(Should only be used in KNN_Mode) Takes a pattern pattern_vector of size vector_length, and uses the committed neurons in the network to classify the pattern

Parameters

  1. uint8_t *pattern_vector : Pointer to the data to be classified. Pattern data must be no longer than 128 bytes
  2. int32_t vector_length : The size (in bytes) of the data to be classified

Return value

CuriePME.noMatch if the input data did not match any of the trained categories. Otherwise, the trained category assigned by the network will be returned