@tensorflow/tfjs-core
Version:
Hardware-accelerated JavaScript library for machine intelligence
113 lines • 5.3 kB
JavaScript
/**
* @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 * as tf from '../index';
import { ALL_ENVS, describeWithFlags } from '../jasmine_util';
import { expectArraysClose } from '../test_util';
describeWithFlags('multinomial', ALL_ENVS, () => {
const NUM_SAMPLES = 1000;
// Allowed Variance in probability (in %).
const EPSILON = 0.05;
const SEED = 3.14;
it('Flip a fair coin and check bounds', async () => {
const probs = tf.tensor1d([1, 1]);
const result = tf.multinomial(probs, NUM_SAMPLES, SEED);
expect(result.dtype).toBe('int32');
expect(result.shape).toEqual([NUM_SAMPLES]);
const outcomeProbs = computeProbs(await result.data(), 2);
expectArraysClose(outcomeProbs, [0.5, 0.5], EPSILON);
});
it('Flip a two-sided coin with 100% of heads', async () => {
const logits = tf.tensor1d([1, -100]);
const result = tf.multinomial(logits, NUM_SAMPLES, SEED);
expect(result.dtype).toBe('int32');
expect(result.shape).toEqual([NUM_SAMPLES]);
const outcomeProbs = computeProbs(await result.data(), 2);
expectArraysClose(outcomeProbs, [1, 0], EPSILON);
});
it('Flip a two-sided coin with 100% of tails', async () => {
const logits = tf.tensor1d([-100, 1]);
const result = tf.multinomial(logits, NUM_SAMPLES, SEED);
expect(result.dtype).toBe('int32');
expect(result.shape).toEqual([NUM_SAMPLES]);
const outcomeProbs = computeProbs(await result.data(), 2);
expectArraysClose(outcomeProbs, [0, 1], EPSILON);
});
it('Flip a single-sided coin throws error', () => {
const probs = tf.tensor1d([1]);
expect(() => tf.multinomial(probs, NUM_SAMPLES, SEED)).toThrowError();
});
it('Flip a ten-sided coin and check bounds', async () => {
const numOutcomes = 10;
const logits = tf.fill([numOutcomes], 1).as1D();
const result = tf.multinomial(logits, NUM_SAMPLES, SEED);
expect(result.dtype).toBe('int32');
expect(result.shape).toEqual([NUM_SAMPLES]);
const outcomeProbs = computeProbs(await result.data(), numOutcomes);
expect(outcomeProbs.length).toBeLessThanOrEqual(numOutcomes);
});
it('Flip 3 three-sided coins, each coin is 100% biases', async () => {
const numOutcomes = 3;
const logits = tf.tensor2d([[-100, -100, 1], [-100, 1, -100], [1, -100, -100]], [3, numOutcomes]);
const result = tf.multinomial(logits, NUM_SAMPLES, SEED);
expect(result.dtype).toBe('int32');
expect(result.shape).toEqual([3, NUM_SAMPLES]);
// First coin always gets last event.
let outcomeProbs = computeProbs((await result.data()).slice(0, NUM_SAMPLES), numOutcomes);
expectArraysClose(outcomeProbs, [0, 0, 1], EPSILON);
// Second coin always gets middle event.
outcomeProbs = computeProbs((await result.data()).slice(NUM_SAMPLES, 2 * NUM_SAMPLES), numOutcomes);
expectArraysClose(outcomeProbs, [0, 1, 0], EPSILON);
// Third coin always gets first event
outcomeProbs =
computeProbs((await result.data()).slice(2 * NUM_SAMPLES), numOutcomes);
expectArraysClose(outcomeProbs, [1, 0, 0], EPSILON);
});
it('passing Tensor3D throws error', () => {
const probs = tf.zeros([3, 2, 2]);
const normalized = true;
expect(() => tf.multinomial(probs, 3, SEED, normalized))
.toThrowError();
});
it('throws when passed a non-tensor', () => {
// tslint:disable-next-line:no-any
expect(() => tf.multinomial({}, NUM_SAMPLES, SEED))
.toThrowError(/Argument 'logits' passed to 'multinomial' must be a Tensor/);
});
it('accepts a tensor-like object for logits (biased coin)', async () => {
const res = tf.multinomial([-100, 1], NUM_SAMPLES, SEED);
expect(res.dtype).toBe('int32');
expect(res.shape).toEqual([NUM_SAMPLES]);
const outcomeProbs = computeProbs(await res.data(), 2);
expectArraysClose(outcomeProbs, [0, 1], EPSILON);
});
function computeProbs(events, numOutcomes) {
const counts = [];
for (let i = 0; i < numOutcomes; ++i) {
counts[i] = 0;
}
const numSamples = events.length;
for (let i = 0; i < events.length; ++i) {
counts[events[i]]++;
}
// Normalize counts to be probabilities between [0, 1].
for (let i = 0; i < counts.length; i++) {
counts[i] /= numSamples;
}
return counts;
}
});
//# sourceMappingURL=multinomial_test.js.map