Model search (MS) is a library that uses ML model architecture automatic search algorithms. The developers claim that the framework scales in cases when the state search space appears large. The framework is based on Bayesian optimization.
The library’s mission is to accelerate the process of finding the optimal architecture for researchers. At the moment MS works only for the classification task with both tabular data and images. MS selects the types of layers in deep neural networks to optimize the error function in the target problem.
More about the library
The Model Search functionality allows you to:
- Run many AutoML algorithms in parallel out of the box on your data. Using MS, you can search for the optimal model architecture, the optimal ensemble of models, and the best-distilled version of the model;
- Compare different models that were found during the search;
- Create your own search space to customize the types of layers in the neural network
A more detailed technical description of the framework’s capabilities is available in the original article.