Computational model optimizations

A non-transitory computer-readable storage medium storing instructions that when executed by a processor circuit cause the processor circuit to:

• receive a plurality of sampled values for a hyperparameter of a computational model, the plurality of sampled values comprising a subset of a plurality of possible values for the hyperparameter, each sampled value associated with a respective performance metric for the computational model with the sampled value assigned to the hyperparameter;
• determine a first candidate value from the plurality of possible values in a search space, wherein the first candidate value is not one of the plurality of sampled values;
• assign the first candidate value to the hyperparameter of the computational model;
• determine a first performance metric for the computational model based on training data and validation data processed by the computational model with the first candidate value assigned to the hyperparameter;
• partition the search space into a plurality of regions;
• determine a first region of the plurality of regions having a lowest count of the plurality of sampled values relative to the remaining regions of the plurality of regions;
• determine a second candidate value in the first region;
• assign the second candidate value to the hyperparameter of the computational model;
• and
• determine a second performance metric for the computational model based on training data and validation data processed by the computational model with the second candidate value assigned to the hyperparameter.

A method, comprising:

• receiving a plurality of sampled values for a hyperparameter of a computational model, the plurality of sampled values comprising a subset of a plurality of possible values for the hyperparameter, each sampled value associated with a respective performance metric for the computational model with the sampled value assigned to the hyperparameter;
• determining a first candidate value from the plurality of possible values in a search space, wherein the first candidate value is not one of the plurality of sampled values;
• assigning the first candidate value to the hyperparameter of the computational model;
• and
• determining, by a processor circuit, a first performance metric for the computational model based on training data and validation data processed by the computational model with the first candidate value assigned to the hyperparameter;
• partitioning the search space into a plurality of regions;
• determining a first region of the plurality of regions having a lowest count of the plurality of sampled values relative to the remaining regions of the plurality of regions;
• determining a second candidate value in the first region;
• assigning the second candidate value to the hyperparameter of the computational model;
• and
• determining a second performance metric for the computational model based on training data and validation data processed by the computational model with the second candidate value assigned to the hyperparameter.