ES6 hyperparameters search for tfjs
APACHE-2.0 License
⚠️ Early version subject to changes.
$ npm install hyperparameters
import * as hpjs from 'hyperparameters';
low
and high
i.e. any value between low
and high
has an equal probability of being selectedhp.uniform
calculated as round(uniform(low, high) / q) * q
exp(uniform(low, high))
so the logarithm of the return value is uniformly distributed.round(exp(uniform(low, high)) / q) * q
round(normal(mu, sigma) / q) * q
exp(normal(mu, sigma))
round(exp(normal(mu, sigma)) / q) * q
import { RandomState } from 'hyperparameters';
example:
const rng = new RandomState(12345);
console.log(rng.randrange(0, 5, 0.5));
import { sample } from 'hyperparameters';
example:
import * as hpjs from 'hyperparameters';
const space = {
x: hpjs.normal(0, 2),
y: hpjs.uniform(0, 1),
choice: hpjs.choice([
undefined, hp.uniform('float', 0, 1),
]),
array: [
hpjs.normal(0, 2), hpjs.uniform(0, 3), hpjs.choice([false, true]),
],
obj: {
u: hpjs.uniform(0, 3),
v: hpjs.uniform(0, 3),
w: hpjs.uniform(-3, 0)
}
};
console.log(hpjs.sample.randomSample(space));
import * as hpjs from 'hyperparameters';
const trials = hpjs.fmin(optimizationFunction, space, estimator, max_estimates, options);
example:
import * as hpjs from 'hyperparameters';
const fn = x => ((x ** 2) - (x + 1));
const space = hpjs.uniform(-5, 5);
fmin(fn, space, hpjs.search.randomSearch, 1000, { rng: new hpjs.RandomState(123456) })
.then(trials => console.log(result.argmin));
<script src="https://cdn.jsdelivr.net/npm/hyperparameters@latest/dist/hyperparameters.min.js" />
const space = {
optimizer: hpjs.choice(['sgd', 'adam', 'adagrad', 'rmsprop']),
epochs: hpjs.quniform(50, 250, 50),
};
const trainModel = async ({ optimizer, epochs }, { xs, ys }) => {
// Create a simple model.
const model = tf.sequential();
model.add(tf.layers.dense({ units: 1, inputShape: [1] }));
// Prepare the model for training: Specify the loss and the optimizer.
model.compile({
loss: 'meanSquaredError',
optimizer
});
// Train the model using the data.
const h = await model.fit(xs, ys, { epochs });
return { model, loss: h.history.loss[h.history.loss.length - 1] };
};
const modelOpt = async ({ optimizer, epochs }, { xs, ys }) => {
const { loss } = await trainModel({ optimizer, epochs }, { xs, ys });
return { loss, status: hpjs.STATUS_OK };
};
const trials = await hpjs.fmin(
modelOpt, space, hpjs.search.randomSearch, 10,
{ rng: new hpjs.RandomState(654321), xs, ys }
);
const opt = trials.argmin;
console.log('best optimizer',opt.optimizer);
console.log('best no of epochs', opt.epochs);
$ npm install hyperparameters
import * as tf from '@tensorflow/tfjs';
import * as hpjs from 'hyperparameters';
const space = {
optimizer: hpjs.choice(['sgd', 'adam', 'adagrad', 'rmsprop']),
epochs: hpjs.quniform(50, 250, 50),
};
const trainModel = async ({ optimizer, epochs }, { xs, ys }) => {
// Create a simple model.
const model = tf.sequential();
model.add(tf.layers.dense({ units: 1, inputShape: [1] }));
// Prepare the model for training: Specify the loss and the optimizer.
model.compile({
loss: 'meanSquaredError',
optimizer
});
// Train the model using the data.
const h = await model.fit(xs, ys, { epochs });
return { model, loss: h.history.loss[h.history.loss.length - 1] };
};
const modelOpt = async ({ optimizer, epochs }, { xs, ys }) => {
const { loss } = await trainModel({ optimizer, epochs }, { xs, ys });
return { loss, status: hpjs.STATUS_OK };
};
const trials = await hpjs.fmin(
modelOpt, space, hpjs.search.randomSearch, 10,
{ rng: new hpjs.RandomState(654321), xs, ys }
);
const opt = trials.argmin;
console.log('best optimizer',opt.optimizer);
console.log('best no of epochs', opt.epochs);
MIT © Atanas Stoyanov & Martin Stoyanov