/**
Copyright (c) 2011, the SemanticVectors AUTHORS.
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following
disclaimer in the documentation and/or other materials provided
with the distribution.
* Neither the name of the University of Pittsburgh nor the names
of its contributors may be used to endorse or promote products
derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
**/
package pitt.search.semanticvectors.vectors;
import java.io.IOException;
import java.util.Arrays;
import java.util.Random;
import org.apache.lucene.store.IOContext;
import org.apache.lucene.store.IndexInput;
import org.apache.lucene.store.IndexOutput;
import org.apache.lucene.store.RAMDirectory;
import org.junit.Test;
import static pitt.search.semanticvectors.MyTestUtils.assertFloatArrayEquals;
import static pitt.search.semanticvectors.vectors.ComplexVector.Mode;
import junit.framework.TestCase;
import static org.junit.Assert.assertArrayEquals;
/**
* Tests for {@code ComplexVector} class. Many of these tests use {@code ComplexVector#toString}
* to check correct results. This is somewhat ad hoc and any change in the debug string
* representation will likely cause breakages. Be warned!
*/
public class ComplexVectorTest extends TestCase {
private static final double TOL = 0.001;
@Test
public void testGenerateRandomVector() {
// Generate random vector
ComplexVector cv1 = (ComplexVector) VectorFactory.generateRandomVector(
VectorType.COMPLEX, 5, 2, new Random(0));
assertArrayEquals(new short[] {0, 13622, 4, 9934}, cv1.getSparseOffsets());
assertEquals(Mode.POLAR_SPARSE, cv1.getOpMode());
}
@Test
public void testDenseToCartesian() {
ComplexVector vector = new ComplexVector(new short[] {-1, 0});
vector.toCartesian();
assertEquals(ComplexVector.Mode.CARTESIAN, vector.getOpMode());
assertFloatArrayEquals(new float[] {0.0f, 0.0f, 1.0f, 0.0f}, vector.getCoordinates(), TOL);
}
@Test
public void testComplexVectorConversion() {
ComplexVector vector = (ComplexVector) VectorFactory.createZeroVector(VectorType.COMPLEX, 10);
assertTrue(vector.isZeroVector());
assertEquals(10, vector.getDimension());
// Note here need to start with a brand new complex vector, not using createZeroVector.
// This is confusing as an API and maybe should be changed.
vector = new ComplexVector(2, Mode.POLAR_SPARSE);
vector.setSparseOffsets(new short[] {1, 0});
vector.toDensePolar();
assertArrayEquals(new short[] {-1, 0}, vector.getPhaseAngles());
vector.toCartesian();
assertFloatArrayEquals(new float[] {0, 0, 1, 0}, vector.getCoordinates(), TOL);
}
@Test
public void testWriteToString() {
float[] coords = { 12.3f, 3.2f, 2.6f, -1.3f, -0.01f, -1000.2f };
ComplexVector cv = new ComplexVector(coords);
assertTrue(cv.writeToString().contains("12.3|3.2|2.6|-1.3|-0.01|-1000.2"));
assertTrue(cv.toString().contains("12.3 3.2 2.6 -1.3 -0.01 -1000.2 "));
}
@Test
public void testSuperposeSparseOnZero() {
int dim = 4;
ComplexVector cv1 = (ComplexVector) VectorFactory.createZeroVector(VectorType.COMPLEX, dim);
cv1.setSparseOffsets(new short[] {1, 0, 3, CircleLookupTable.PHASE_RESOLUTION / 4});
ComplexVector cv2 = (ComplexVector) VectorFactory.createZeroVector(VectorType.COMPLEX, dim);
cv2.superpose(cv1, 5, null);
assertEquals(Mode.CARTESIAN, cv2.getOpMode());
assertFloatArrayEquals(
new float[] {0, 0, 5, 0, 0, 0, 0, 5}, cv2.getCoordinates(), TOL);
}
@Test
public void testSuperposeZeroOnSparse() {
int dim = 4;
int seedLength = 2;
ComplexVector cv1 = (ComplexVector) VectorFactory.generateRandomVector(
VectorType.COMPLEX, dim, seedLength, new Random(0));
assertArrayEquals(new short[] {2, 13622, 0, 9934}, cv1.getSparseOffsets());
ComplexVector cv2 = (ComplexVector) VectorFactory.createZeroVector(VectorType.COMPLEX, dim);
cv2.superpose(cv1, 1.0, null);
assertFloatArrayEquals(
new float[] {-0.78503f, -0.61945f, 0, 0, 0.48955f, -0.872064f, 0, 0},
cv2.getCoordinates(), TOL);
}
@Test
public void testVectorCopyCartesian() {
float[] coords = { 12.3f, 3.2f, 2.6f, -1.3f };
ComplexVector cv1 = new ComplexVector(coords);
ComplexVector cv2 = cv1.copy();
assertFloatArrayEquals(
new float[] { 12.3f, 3.2f, 2.6f, -1.3f }, cv2.getCoordinates(), TOL);
}
@Test
public void testCartesianToDensePolar() {
ComplexVector cv = new ComplexVector( new float[] {1.0f, 1.0f, 0, 1.0f, 1.0f, 0 } );
cv.toDensePolar();
assertArrayEquals(new short[] {2048, 4096, 0}, cv.getPhaseAngles());
}
@Test
public void testNormalize() {
ComplexVector.setDominantMode(Mode.CARTESIAN);
ComplexVector cv = new ComplexVector(new float[] {3, 4, 4, 3, 4, 3, 0, 5});
cv.normalize();
System.out.println(cv.toString());
// Note that normalize() makes each individual pair give a coordinate on the unit circle.
assertFloatArrayEquals(new float[] {0.6f, 0.8f, 0.8f, 0.6f, 0.8f, 0.6f, 0.0f, 1.0f},
cv.getCoordinates(), TOL);
ComplexVector.setDominantMode(Mode.POLAR_DENSE);
cv.normalize();
assertArrayEquals(new short[] {2417, 1676, 1676, 4096}, cv.getPhaseAngles());
}
@Test
public void testMeasurePolarOverlap() {
int RES = CircleLookupTable.PHASE_RESOLUTION;
short[] angles1 = { 0, 0, 0 };
short[] angles2 = { 0, -1, (short) (RES / 4) }; // Remember that -1 angle means complex zero.
short[] angles3 = { (short) (3*RES / 4), (short) (RES / 2), 0 };
ComplexVector cv1 = new ComplexVector(angles1);
ComplexVector cv2 = new ComplexVector(angles2);
ComplexVector cv3 = new ComplexVector(angles3);
assertEquals(1/3f, cv1.measurePolarDenseOverlap(cv2), TOL);
assertEquals(0, cv1.measurePolarDenseOverlap(cv3), TOL);
assertEquals(0, cv2.measurePolarDenseOverlap(cv3), TOL);
assertEquals(1, cv1.measurePolarDenseOverlap(cv1), TOL);
assertEquals(1, cv2.measurePolarDenseOverlap(cv2), TOL); // Zero entry doesn't contribute.
assertEquals(1, cv3.measurePolarDenseOverlap(cv3), TOL);
}
@Test
public void testConvolve() {
short RES = CircleLookupTable.PHASE_RESOLUTION;
short ZERO_INDEX = CircleLookupTable.ZERO_INDEX;
ComplexVector cv = new ComplexVector(new short[] {ZERO_INDEX, 0, (short) (RES/2)});
ComplexVector cv2 = new ComplexVector(new short[] {0, (short) (RES/4), (short) (RES/4)});
cv.convolve(cv2, 1);
for (short i : cv.getPhaseAngles()) {
System.err.print(i + " ");
}
assertArrayEquals(new short[] {0, (short) (RES/4), (short) (3*RES/4)}, cv.getPhaseAngles());
// Should have what we started with.
cv.convolve(cv2, -1);
assertArrayEquals(new short[] {0, 0, (short) (RES/2)}, cv.getPhaseAngles());
// Convolving with inverse of itself gives all ones (or zeros).
cv.convolve(cv, -1);
assertArrayEquals(new short[] {0, 0, 0}, cv.getPhaseAngles());
}
@Test
public void testBindFromRandom() {
short ZERO_INDEX = CircleLookupTable.ZERO_INDEX;
Random random = new Random(0);
ComplexVector cv1 = (ComplexVector) VectorFactory.generateRandomVector(
VectorType.COMPLEX, 5, 2, random);
ComplexVector cv2 = (ComplexVector) VectorFactory.generateRandomVector(
VectorType.COMPLEX, 5, 2, random);
cv1.bind(cv2);
assertArrayEquals(new short[] {2301, 1917, ZERO_INDEX, ZERO_INDEX, 9934}, cv1.getPhaseAngles());
}
@Test
public void testBindFromZero() {
short ZERO_INDEX = CircleLookupTable.ZERO_INDEX;
Random random = new Random(0);
ComplexVector cv1 = (ComplexVector) VectorFactory.createZeroVector(VectorType.COMPLEX, 5);
cv1.toDensePolar();
ComplexVector cv2 = (ComplexVector) VectorFactory.generateRandomVector(
VectorType.COMPLEX, 5, 2, random);
cv1.bind(cv2);
assertArrayEquals(
new short[] {13622, ZERO_INDEX, ZERO_INDEX, ZERO_INDEX, 9934}, cv1.getPhaseAngles());
}
@Test
public void testComplexOrthogonalize() {
short ZERO_INDEX = CircleLookupTable.ZERO_INDEX;
ComplexVector cv = new ComplexVector(new short[] {ZERO_INDEX, 0});
ComplexVector cv2 = new ComplexVector(new short[] {0, 0});
ComplexVectorUtils.renderPairwiseOrthogonal(cv2, cv);
cv2.normalize();
assertArrayEquals(new short[] {0, ZERO_INDEX}, cv2.getPhaseAngles());
}
@Test
public void testReadWrite() {
Vector v1 = new ComplexVector(new short[] { -1, 8000, 16000 });
RAMDirectory directory = new RAMDirectory();
try {
IndexOutput indexOutput = directory.createOutput("complexvectors.bin", IOContext.DEFAULT);
v1.writeToLuceneStream(indexOutput);
indexOutput.close();
IndexInput indexInput = directory.openInput("complexvectors.bin", IOContext.DEFAULT);
ComplexVector cv2 = new ComplexVector(3, Mode.POLAR_SPARSE);
cv2.readFromLuceneStream(indexInput);
assertFloatArrayEquals(
new float[] {0, 0, -0.997290f, 0.073564f, 0.989176f, -0.1467304f},
cv2.getCoordinates(), TOL);
} catch (IOException e) {
e.printStackTrace();
fail();
}
directory.close();
}
}
