Thank you for presenting your work! You note that in Info-Metrics given a set of models with similar performance you choose the simplest model. It seems (correct me if I'm mischaracterizing) that 'simpler' corresponds to fewer parameters to estimate. I was wondering if a simple comparison of parameter count captures the entire complexity of a model? I feel like given two models with an equal parameter count they might not search over the equivalent function space. Thanks!
Thank you for presenting your work! You note that in Info-Metrics given a set of models with similar performance you choose the simplest model. It seems (correct me if I'm mischaracterizing) that 'simpler' corresponds to fewer parameters to estimate. I was wondering if a simple comparison of parameter count captures the entire complexity of a model? I feel like given two models with an equal parameter count they might not search over the equivalent function space. Thanks!