Analysis device, analysis method, and analysis program

Abstract

Provided are an analysis device, an analysis method, and an analysis program which predict the performance of a learning model when learning processing is executed using multiple algorithms. Using a predictive model produced by supervised learning using first shape information representing a global shape of a first loss function set for a prescribed problem and the performance of the learning model as learning data, an analysis device 10 predicts, for each of the multiple algorithms, the performance of a learning model when machine learning by the learning model is executed so that a second loss function has a reduced value on the basis of second shape information representing a global shape of the second loss function set for a new problem.

What is claimed is: 
     
         1 . An analysis device, comprising:
 a learning unit which optimizes one or more hyperparameters included in each of multiple algorithms using multiple optimization algorithms and performs machine learning by a prescribed learning model using the multiple algorithms including the one or more optimized hyperparameters; and   a computing unit which computes the performance of the learning model on the basis of the machine learning, for each of the algorithms and the optimization algorithms.   
     
     
         2 . The analysis device according to  claim 1 , wherein the algorithm is a reinforcement learning algorithm,
 the learning unit performs machine learning by a prescribed learning model on the basis of training data set for a prescribed problem using a plurality of the reinforcement learning algorithms, and   the computing unit computes the performance of the learning model using test data set for the prescribed problem.   
     
     
         3 . The analysis device according to  claim 2 , further comprising an estimation model producing unit which performs supervised learning using, as learning data, the training data and a combination of an optimization algorithm and a reinforcement learning algorithm selected on the basis of the performance and produces an estimation model for estimating a combination of an optimization algorithm and a reinforcement learning algorithm corresponding to a new problem. 
     
     
         4 . An analysis method causing a processor provided in an analysis device to carry out:
 optimizing one or more hyperparameters included in each of multiple algorithms using multiple optimization algorithms and performing machine learning by a prescribed learning model using the multiple algorithms including the one or more optimized hyperparameters; and   computing the performance of the learning model, for each of the algorithms and the optimization algorithms.   
     
     
         5 . An analysis program causing a processor provided in an analysis device to carry out:
 optimizing one or more hyperparameters included in each of multiple algorithms using multiple optimization algorithms, and performing machine learning by a prescribed learning model using the multiple algorithms including the one or more optimized hyperparameters; and   computing the performance of the learning model for each of the algorithms and the optimization algorithms.