org.numenta.nupic.encoders.Encoder Java Examples

/**
 * Stores a {@link NamedTuple} which contains the input values and bucket
 * information - keyed to the encoded field name so that a classifier can
 * retrieve it later on in the processing sequence.
 * 
 * @param inference
 * @param encoderInputMap
 */
private void doEncoderBucketMapping(Inference inference, Map<String, Object> encoderInputMap) {
    if(encoderTuples == null) {
        encoderTuples = encoder.getEncoders(encoder);
    }

    // Store the encoding
    int[] encoding = inference.getEncoding();

    for(EncoderTuple t : encoderTuples) {
        String name = t.getName();
        Encoder<?> e = t.getEncoder();

        int bucketIdx = -1;
        Object o = encoderInputMap.get(name);
        if(DateTime.class.isAssignableFrom(o.getClass())) {
            bucketIdx = ((DateEncoder)e).getBucketIndices((DateTime)o)[0];
        } else if(Number.class.isAssignableFrom(o.getClass())) {
            bucketIdx = e.getBucketIndices((double)o)[0];
        } else {
            bucketIdx = e.getBucketIndices((String)o)[0];
        }

        int offset = t.getOffset();
        int[] tempArray = new int[e.getWidth()];
        System.arraycopy(encoding, offset, tempArray, 0, tempArray.length);

        inference.getClassifierInput().put(
                name,
                new NamedTuple(
                        new String[] { "name", "inputValue", "bucketIdx", "encoding" },
                        name,
                        o,
                        bucketIdx,
                        tempArray
                ));
    }
}
 

/**
 * Searches through the specified {@link MultiEncoder}'s previously configured 
 * encoders to find and return one that is of type {@link CoordinateEncoder} or
 * {@link GeospatialCoordinateEncoder}
 * 
 * @param enc   the containing {@code MultiEncoder}
 * @return
 */
private Optional<Encoder<?>> getCoordinateEncoder(MultiEncoder enc) {
    for(EncoderTuple t : enc.getEncoders(enc)) {
        if((t.getEncoder() instanceof CoordinateEncoder) ||
            (t.getEncoder() instanceof GeospatialCoordinateEncoder)) {
            return Optional.of(t.getEncoder());
        }
    }
    
    return Optional.empty();
}
 

/**
 * Searches through the specified {@link MultiEncoder}'s previously configured 
 * encoders to find and return one that is of type {@link CategoryEncoder} or
 * {@link SDRCategoryEncoder}
 * 
 * @param enc   the containing {@code MultiEncoder}
 * @return
 */
private Optional<Encoder<?>> getCategoryEncoder(MultiEncoder enc) {
    for(EncoderTuple t : enc.getEncoders(enc)) {
        if((t.getEncoder() instanceof CategoryEncoder) ||
            (t.getEncoder() instanceof SDRCategoryEncoder)) {
            return Optional.of(t.getEncoder());
        }
    }
    
    return Optional.empty();
}
 

/**
 * Searches through the specified {@link MultiEncoder}'s previously configured 
 * encoders to find and return one that is of type {@link ScalarEncoder},
 * {@link RandomDistributedScalarEncoder}, {@link AdaptiveScalarEncoder},
 * {@link LogEncoder} or {@link DeltaEncoder}.
 * 
 * @param enc   the containing {@code MultiEncoder}
 * @return
 */
private Optional<Encoder<?>> getNumberEncoder(MultiEncoder enc) {
    for(EncoderTuple t : enc.getEncoders(enc)) {
        if((t.getEncoder() instanceof RandomDistributedScalarEncoder) ||
            (t.getEncoder() instanceof ScalarEncoder) ||
            (t.getEncoder() instanceof AdaptiveScalarEncoder) ||
            (t.getEncoder() instanceof LogEncoder) ||
            (t.getEncoder() instanceof DeltaEncoder)) {
            
            return Optional.of(t.getEncoder());
        }
    }
    
    return Optional.empty();
 }
 

/**
 * Sets up a mapping which describes the order of occurrence of comma
 * separated fields - mapping their ordinal position to the {@link Encoder}
 * which services the encoding of the field occurring in that position. This
 * sequence of types is contained by an instance of {@link Header} which
 * makes available an array of {@link FieldMetaType}s.
 */
private void makeIndexEncoderMap() {
    indexToEncoderMap = new TIntObjectHashMap<Encoder<?>>();
    
    for (int i = 0, size = header.getFieldNames().size(); i < size; i++) {
        switch (header.getFieldTypes().get(i)) {
            case DATETIME:
                Optional<DateEncoder> de = getDateEncoder(encoder);
                if (de.isPresent()) {
                    indexToEncoderMap.put(i, de.get());
                } else {
                    throw new IllegalArgumentException("DateEncoder never initialized: " + header.getFieldNames().get(i));
                }
                break;
            case BOOLEAN:
            case FLOAT:
            case INTEGER:
                Optional<Encoder<?>> ne = getNumberEncoder(encoder);
                if (ne.isPresent()) {
                    indexToEncoderMap.put(i, ne.get());
                } else {
                    throw new IllegalArgumentException("Number (or Boolean) encoder never initialized: " + header.getFieldNames().get(i));
                }
                break;
            case LIST:
            case STRING:
                Optional<Encoder<?>> ce = getCategoryEncoder(encoder);
                if (ce.isPresent()) {
                    indexToEncoderMap.put(i, ce.get());
                } else {
                    throw new IllegalArgumentException("Category encoder never initialized: " + header.getFieldNames().get(i));
                }
                break;
            case COORD:
            case GEO:
                Optional<Encoder<?>> ge = getCoordinateEncoder(encoder);
                if (ge.isPresent()) {
                    indexToEncoderMap.put(i, ge.get());
                } else {
                    throw new IllegalArgumentException("Coordinate encoder never initialized: " + header.getFieldNames().get(i));
                }
                break;
            case SARR:
            case DARR:
                Optional<SDRPassThroughEncoder> spte = getSDRPassThroughEncoder(encoder);
                if (spte.isPresent()) {
                    indexToEncoderMap.put(i, spte.get());
                } else {
                    throw new IllegalArgumentException("SDRPassThroughEncoder encoder never initialized: " + header.getFieldNames().get(i));
                }
                break;
            default:
                break;
        }
    }
    
}
 

/**
 * Special test case for extra processing for category encoder lists
 */
@Test
public void testCategoryEncoderCreation() {
    Sensor<File> sensor = Sensor.create(
        FileSensor::create, SensorParams.create(
            Keys::path, "", ResourceLocator.path("rec-center-hourly-4period-cat.csv")));
    
    
    // Cast the ValueList to the more complex type (Header)
    HTMSensor<File> htmSensor = (HTMSensor<File>)sensor;
    Header meta = (Header)htmSensor.getMetaInfo();
    assertTrue(meta.getFieldTypes().stream().allMatch(
        l -> l.equals(FieldMetaType.DATETIME) || l.equals(FieldMetaType.FLOAT) || l.equals(FieldMetaType.LIST)));
    assertTrue(meta.getFieldNames().stream().allMatch(
        l -> l.equals("timestamp") || l.equals("consumption") || l.equals("type")));
    assertTrue(meta.getFlags().stream().allMatch(
        l -> l.equals(SensorFlags.T) || l.equals(SensorFlags.B) || l.equals(SensorFlags.C)));
    
    // Set the parameters on the sensor.
    // This enables it to auto-configure itself; a step which will
    // be done at the Region level.
    Encoder<Object> multiEncoder = htmSensor.getEncoder();
    assertNotNull(multiEncoder);
    assertTrue(multiEncoder instanceof MultiEncoder);
    
    // Set the Local parameters on the Sensor
    htmSensor.initEncoder(getCategoryEncoderParams());
    List<EncoderTuple> encoders = multiEncoder.getEncoders(multiEncoder);
    assertEquals(3, encoders.size());
    
    DateEncoder dateEnc = (DateEncoder)encoders.get(1).getEncoder();
    SDRCategoryEncoder catEnc = (SDRCategoryEncoder)encoders.get(2).getEncoder();
    assertEquals("[0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]", Arrays.toString(catEnc.encode("ES")));
    
    // Now test the encoding of an input row
    Map<String, Object> d = new HashMap<String, Object>();
    d.put("timestamp", dateEnc.parse("7/12/10 13:10"));
    d.put("consumption", 35.3);
    d.put("type", "ES");
    int[] output = multiEncoder.encode(d);
    System.out.println("output = "+ Arrays.toString(output));
    int[] expected = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 
                      0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
    assertTrue(Arrays.equals(expected, output));
}
 

/**
 * Tests the auto-creation of Encoders from Sensor meta data.
 */
@Test
public void testInternalEncoderCreation() {
    
    Sensor<File> sensor = Sensor.create(
        FileSensor::create, 
        SensorParams.create(
            Keys::path, "", ResourceLocator.path("rec-center-hourly.csv")));
    
    
    // Cast the ValueList to the more complex type (Header)
    HTMSensor<File> htmSensor = (HTMSensor<File>)sensor;
    Header meta = (Header)htmSensor.getMetaInfo();
    assertTrue(meta.getFieldTypes().stream().allMatch(
        l -> l.equals(FieldMetaType.DATETIME) || l.equals(FieldMetaType.FLOAT)));
    assertTrue(meta.getFieldNames().stream().allMatch(
        l -> l.equals("timestamp") || l.equals("consumption")));
    assertTrue(meta.getFlags().stream().allMatch(
        l -> l.equals(SensorFlags.T) || l.equals(SensorFlags.B)));
    
    // Set the parameters on the sensor.
    // This enables it to auto-configure itself; a step which will
    // be done at the Region level.
    Encoder<Object> multiEncoder = htmSensor.getEncoder();
    assertNotNull(multiEncoder);
    assertTrue(multiEncoder instanceof MultiEncoder);
    
    // Set the Local parameters on the Sensor
    htmSensor.initEncoder(getTestEncoderParams());
    List<EncoderTuple> encoders = multiEncoder.getEncoders(multiEncoder);
    assertEquals(2, encoders.size());
    
    // Test date specific encoder configuration
    //
    // All encoders in the MultiEncoder are accessed in a particular
    // order (the alphabetical order their corresponding fields are in),
    // so alphabetically "consumption" proceeds "timestamp"
    // so we need to ensure that the proper order is preserved (i.e. exists at index 1)
    DateEncoder dateEnc = (DateEncoder)encoders.get(1).getEncoder();
    try {
        dateEnc.parseEncode("7/12/10 13:10");
        dateEnc.parseEncode("7/12/2010 13:10");
        // Should fail here due to conflict with configured format
        dateEnc.parseEncode("13:10 7/12/10");
        fail();
    }catch(Exception e) {
       assertEquals("Invalid format: \"13:10 7/12/10\" is malformed at \":10 7/12/10\"", e.getMessage());
    }
    
    RandomDistributedScalarEncoder rdse = (RandomDistributedScalarEncoder)encoders.get(0).getEncoder();
    int[] encoding = rdse.encode(35.3);
    System.out.println(Arrays.toString(encoding));
    
    // Now test the encoding of an input row
    Map<String, Object> d = new HashMap<String, Object>();
    d.put("timestamp", dateEnc.parse("7/12/10 13:10"));
    d.put("consumption", 35.3);
    int[] output = multiEncoder.encode(d);
    int[] expected = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 
                      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
    assertTrue(Arrays.equals(expected, output));
}