The Living Thing / Notebooks :

AutoML

hyperparameter selection with the use of yet more hyperparameters

The sub-field of optimisation that specifically aims to automate model selection in machine learning. (and also occasionally ensemble construction)

Quoc Le & Barret Zoph weigh in for google:

Typically, our machine learning models are painstakingly designed by a team of engineers and scientists. This process of manually designing machine learning models is difficult because the search space of all possible models can be combinatorially large — a typical 10-layer network can have ~1010 candidate networks! […]

To make this process of designing machine learning models much more accessible, we’ve been exploring ways to automate the design of machine learning models. […] in this blog post, we’ll focus on our reinforcement learning approach and the early results we’ve gotten so far.

In our approach (which we call “AutoML”), a controller neural net can propose a “child” model architecture, which can then be trained and evaluated for quality on a particular task. That feedback is then used to inform the controller how to improve its proposals for the next round.

Should you bother getting fancy about this? Ben Recht argues no, that random search is competitive with highly tuned Bayesian methods in hyperparameter tuning. Let's ignore him for a moment though and sniff in the hype.

Differentiable hyperparameter optimisation

MaDA15:

Hyperparameter optimization by gradient descent

Each meta-iteration runs an entire training run of stochastic gradient de- scent to optimize elementary parameters (weights 1 and 2). Gradients of the validation loss with respect to hyperparameters are then computed by propagating gradients back through the elementary training iterations. Hyperparameters (in this case, learning rate and momentum schedules) are then updated in the direction of this hypergradient. … The last remaining parameter to SGD is the initial parameter vector. Treating this vector as a hyperparameter blurs the distinction between learning and meta-learning. In the extreme case where all elementary learning rates are set to zero, the training set ceases to matter and the meta-learning procedure exactly reduces to elementary learning on the validation set. Due to philosophical vertigo, we chose not to optimize the initial parameter vector.

Their implementation, hypergrad, is cool, but no longer maintained.

Bayesian/surrogate optimisation

See Bayesian optimisation

Implementations

Refs