@tensorflow/tfjs-core/dist/ops/rand_test.js

/**
 * @license
 * Copyright 2017 Google LLC. All Rights Reserved.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * =============================================================================
 */
import { util } from '..';
import * as tf from '../index';
import { ALL_ENVS, describeWithFlags } from '../jasmine_util';
import { expectValuesInRange } from '../test_util';
import { MPRandGauss, RandGamma, UniformRandom } from './rand_util';
import { expectArrayInMeanStdRange, jarqueBeraNormalityTest } from './rand_util';
describeWithFlags('rand', ALL_ENVS, () => {
    it('should return a random 1D float32 array', async () => {
        const shape = [10];
        // Enusre defaults to float32 w/o type:
        let result = tf.rand(shape, () => util.randUniform(0, 2));
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 2);
        result = tf.rand(shape, () => util.randUniform(0, 1.5));
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 1.5);
    });
    it('should return a random 1D int32 array', async () => {
        const shape = [10];
        const result = tf.rand(shape, () => util.randUniform(0, 2), 'int32');
        expect(result.dtype).toBe('int32');
        expectValuesInRange(await result.data(), 0, 2);
    });
    it('should return a random 1D bool array', async () => {
        const shape = [10];
        const result = tf.rand(shape, () => util.randUniform(0, 1), 'bool');
        expect(result.dtype).toBe('bool');
        expectValuesInRange(await result.data(), 0, 1);
    });
    it('should return a random 2D float32 array', async () => {
        const shape = [3, 4];
        // Enusre defaults to float32 w/o type:
        let result = tf.rand(shape, () => util.randUniform(0, 2.5));
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 2.5);
        result = tf.rand(shape, () => util.randUniform(0, 1.5), 'float32');
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 1.5);
    });
    it('should return a random 2D int32 array', async () => {
        const shape = [3, 4];
        const result = tf.rand(shape, () => util.randUniform(0, 2), 'int32');
        expect(result.dtype).toBe('int32');
        expectValuesInRange(await result.data(), 0, 2);
    });
    it('should return a random 2D bool array', async () => {
        const shape = [3, 4];
        const result = tf.rand(shape, () => util.randUniform(0, 1), 'bool');
        expect(result.dtype).toBe('bool');
        expectValuesInRange(await result.data(), 0, 1);
    });
    it('should return a random 3D float32 array', async () => {
        const shape = [3, 4, 5];
        // Enusre defaults to float32 w/o type:
        let result = tf.rand(shape, () => util.randUniform(0, 2.5));
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 2.5);
        result = tf.rand(shape, () => util.randUniform(0, 1.5), 'float32');
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 1.5);
    });
    it('should return a random 3D int32 array', async () => {
        const shape = [3, 4, 5];
        const result = tf.rand(shape, () => util.randUniform(0, 2), 'int32');
        expect(result.dtype).toBe('int32');
        expectValuesInRange(await result.data(), 0, 2);
    });
    it('should return a random 3D bool array', async () => {
        const shape = [3, 4, 5];
        const result = tf.rand(shape, () => util.randUniform(0, 1), 'bool');
        expect(result.dtype).toBe('bool');
        expectValuesInRange(await result.data(), 0, 1);
    });
    it('should return a random 4D float32 array', async () => {
        const shape = [3, 4, 5, 6];
        // Enusre defaults to float32 w/o type:
        let result = tf.rand(shape, () => util.randUniform(0, 2.5));
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 2.5);
        result = tf.rand(shape, () => util.randUniform(0, 1.5));
        expect(result.dtype).toBe('float32');
        expectValuesInRange(await result.data(), 0, 1.5);
    });
    it('should return a random 4D int32 array', async () => {
        const shape = [3, 4, 5, 6];
        const result = tf.rand(shape, () => util.randUniform(0, 2), 'int32');
        expect(result.dtype).toBe('int32');
        expectValuesInRange(await result.data(), 0, 2);
    });
    it('should return a random 4D bool array', async () => {
        const shape = [3, 4, 5, 6];
        const result = tf.rand(shape, () => util.randUniform(0, 1), 'bool');
        expect(result.dtype).toBe('bool');
        expectValuesInRange(await result.data(), 0, 1);
    });
});
function isFloat(n) {
    return Number(n) === n && n % 1 !== 0;
}
describe('MPRandGauss', () => {
    const EPSILON = 0.05;
    const SEED = 2002;
    it('should default to float32 numbers', () => {
        const rand = new MPRandGauss(0, 1.5);
        expect(isFloat(rand.nextValue())).toBe(true);
    });
    it('should handle a mean/stdv of float32 numbers', () => {
        const rand = new MPRandGauss(0, 1.5, 'float32', false /* truncated */, SEED);
        const values = [];
        const size = 10000;
        for (let i = 0; i < size; i++) {
            values.push(rand.nextValue());
        }
        expectArrayInMeanStdRange(values, 0, 1.5, EPSILON);
        jarqueBeraNormalityTest(values);
    });
    it('should handle int32 numbers', () => {
        const rand = new MPRandGauss(0, 1, 'int32');
        expect(isFloat(rand.nextValue())).toBe(false);
    });
    it('should handle a mean/stdv of int32 numbers', () => {
        const rand = new MPRandGauss(0, 2, 'int32', false /* truncated */, SEED);
        const values = [];
        const size = 10000;
        for (let i = 0; i < size; i++) {
            values.push(rand.nextValue());
        }
        expectArrayInMeanStdRange(values, 0, 2, EPSILON);
        jarqueBeraNormalityTest(values);
    });
    it('Should not have a more than 2x std-d from mean for truncated values', () => {
        const stdv = 1.5;
        const rand = new MPRandGauss(0, stdv, 'float32', true /* truncated */);
        for (let i = 0; i < 1000; i++) {
            expect(Math.abs(rand.nextValue())).toBeLessThan(stdv * 2);
        }
    });
});
describe('RandGamma', () => {
    const SEED = 2002;
    it('should default to float32 numbers', () => {
        const rand = new RandGamma(2, 2, 'float32');
        expect(isFloat(rand.nextValue())).toBe(true);
    });
    it('should handle an alpha/beta of float32 numbers', () => {
        const rand = new RandGamma(2, 2, 'float32', SEED);
        const values = [];
        const size = 10000;
        for (let i = 0; i < size; i++) {
            values.push(rand.nextValue());
        }
        expectValuesInRange(values, 0, 30);
    });
    it('should handle int32 numbers', () => {
        const rand = new RandGamma(2, 2, 'int32');
        expect(isFloat(rand.nextValue())).toBe(false);
    });
    it('should handle an alpha/beta of int32 numbers', () => {
        const rand = new RandGamma(2, 2, 'int32', SEED);
        const values = [];
        const size = 10000;
        for (let i = 0; i < size; i++) {
            values.push(rand.nextValue());
        }
        expectValuesInRange(values, 0, 30);
    });
});
describe('UniformRandom', () => {
    it('float32, no seed', () => {
        const min = 0.2;
        const max = 0.24;
        const dtype = 'float32';
        const xs = [];
        for (let i = 0; i < 10; ++i) {
            const rand = new UniformRandom(min, max, dtype);
            const x = rand.nextValue();
            xs.push(x);
        }
        expect(Math.min(...xs)).toBeGreaterThanOrEqual(min);
        expect(Math.max(...xs)).toBeLessThan(max);
    });
    it('int32, no seed', () => {
        const min = 13;
        const max = 37;
        const dtype = 'int32';
        const xs = [];
        for (let i = 0; i < 10; ++i) {
            const rand = new UniformRandom(min, max, dtype);
            const x = rand.nextValue();
            expect(Number.isInteger(x)).toEqual(true);
            xs.push(x);
        }
        expect(Math.min(...xs)).toBeGreaterThanOrEqual(min);
        expect(Math.max(...xs)).toBeLessThanOrEqual(max);
    });
    it('seed is number', () => {
        const min = -1.2;
        const max = -0.4;
        const dtype = 'float32';
        const seed = 1337;
        const xs = [];
        for (let i = 0; i < 10; ++i) {
            const rand = new UniformRandom(min, max, dtype, seed);
            const x = rand.nextValue();
            expect(x).toBeGreaterThanOrEqual(min);
            expect(x).toBeLessThan(max);
            xs.push(x);
        }
        // Assert deterministic results.
        expect(Math.min(...xs)).toEqual(Math.max(...xs));
    });
    it('seed === null', () => {
        const min = 0;
        const max = 1;
        const dtype = 'float32';
        const seed = null;
        const rand = new UniformRandom(min, max, dtype, seed);
        const x = rand.nextValue();
        expect(x).toBeGreaterThanOrEqual(0);
        expect(x).toBeLessThan(1);
    });
    it('seed === undefined', () => {
        const min = 0;
        const max = 1;
        const dtype = 'float32';
        const seed = undefined;
        const rand = new UniformRandom(min, max, dtype, seed);
        const x = rand.nextValue();
        expect(x).toBeGreaterThanOrEqual(0);
        expect(x).toBeLessThan(1);
    });
});
//# sourceMappingURL=rand_test.js.map