Google Model Search — AI finds automatically the best architecture

An improvement upon AutoML is here!

Google has just announced Model Search — a framework that implements AutoML algorithms for model architecture search (at scale).

The goal here is to make the exploration and discovery process of the right ML architecture much faster. Let’s dive into it!

Google Model Search — AutoML framework

How to use AutoML from Google

The simplest case is described in Model Search Github documentation.

We start with a csv file where the features are numbers and we want AutoML find us the best model architecture.

This is how to do it:

import model_search
from model_search import constants
from model_search import single_trainer
from model_search.data import csv_data
trainer = single_trainer.SingleTrainer(
    data=csv_data.Provider(
        label_index=0,
        logits_dimension=2,
        record_defaults=[0, 0, 0, 0],
        filename="model_search/data/testdata/csv_random_data.csv")),
    spec=constants.DEFAULT_DNN)
trainer.try_models(
    number_models=200,
    train_steps=1000,
    eval_steps=100,
    root_dir="/tmp/run_example",
    batch_size=32,
    experiment_name="example"…