/**
|
* @license
|
* Copyright 2018 Google Inc. 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 {Scalar, Tensor} from '../tensor';
|
import {assertTypesMatch} from '../tensor_util';
|
import {convertToTensor} from '../tensor_util_env';
|
import {TensorLike} from '../types';
|
import * as util from '../util';
|
import {pow} from './binary_ops';
|
import {op} from './operation';
|
import {scalar} from './tensor_ops';
|
|
/**
|
* Compute the moving average of a variable.
|
*
|
* Without zeroDebias, the moving average operation is defined by:
|
* `v += delta`
|
* where
|
* `delta = (1 - decay) * (x - v)`
|
*
|
* With zeroDebias (default), the `delta` term is scaled to debias the
|
* effect of the (assumed) zero-initialization of `v`.
|
* `delta /= (1 - decay ^ step)`
|
*
|
* For more details on the zero-debiasing algorithm, see:
|
* https://arxiv.org/abs/1412.6980
|
*
|
* Note that this function is completely stateless and does not keep track of
|
* step count. The step count needs to be maintained by the caller and passed
|
* in as `step`.
|
*
|
* @param v The current moving average value.
|
* @param x New input value, must have the same shape and dtype as `v`.
|
* @param decay The decay factor. Typical values are 0.95 and 0.99.
|
* @param step Step count.
|
* @param zeroDebias: Whether zeroDebias is to be performed (default: `true`).
|
* @returns The new moving average value.
|
*/
|
/** @doc {heading: 'Operations', subheading: 'Moving Average'} */
|
function movingAverage_<T extends Tensor>(
|
v: T|TensorLike, x: T|TensorLike, decay: number|Scalar,
|
step?: number|Scalar, zeroDebias = true): T {
|
const $v = convertToTensor(v, 'v', 'movingAverage');
|
const $x = convertToTensor(x, 'x', 'movingAverage');
|
const $decay = convertToTensor(decay, 'decay', 'movingAverage');
|
|
assertTypesMatch($v, $x);
|
util.assert(
|
util.arraysEqual($v.shape, $x.shape), () => 'Shape mismatch in v and x');
|
|
const one = scalar(1);
|
const oneMinusDecay = one.sub($decay);
|
|
let update = $x.sub($v).mul(oneMinusDecay);
|
if (zeroDebias) {
|
util.assert(
|
step != null, () => 'When using zeroDebias: true, step is required.');
|
const $step = convertToTensor(step, 'step', 'movingAverage');
|
update = update.div(one.sub(pow($decay, $step)));
|
}
|
return $v.add(update);
|
}
|
|
export const movingAverage = op({movingAverage_});
|