Genetic programming for partial layers of a deep learning model
An information processing device according to one embodiment includes a processor. The processor executes an acquisition step of acquiring a plurality of partial layers as a gene pool, the partial layers being candidates for elements of a deep learning model, a generation step of generating a new partial layer from the gene pool by using genetic programming, an evaluation step of evaluating each partial layer by incorporating, into a template of the deep learning model, each partial layer in the gene pool to which the new partial layer is added, and determining a plurality of partial layers to remain in the gene pool, and an output step of outputting the deep learning model into which the partial layer with a highest evaluation value is incorporated.
1. An information processing device comprising:
at least one memory operable to store program code; and
at least one processor operable to read the program code and operate as instructed by the program code, the program code being configured to cause the at least one processor to:
acquire a plurality of partial layers as a gene pool, the partial layers being candidates for elements of a deep learning model;
generate a new partial layer from the gene pool by using genetic programming;
evaluate each partial layer by incorporating, into a template of the deep learning model, each partial layer in the gene pool to which the new partial layer is added, and determine a plurality of partial layers to remain in the gene pool; and
output the deep learning model into which the partial layer with a highest evaluation value is incorporated.
2. The information processing device according to claim 1 , wherein
the program code is further configured to cause the at least one processor to output the deep learning model after repeating a search including the generation and the evaluation.
3. The information processing device according to claim 2 , wherein
the program code is further configured to cause the at least one processor to repeat the search n number of times, where n is a natural number greater than 1, and
the program code is further configured to cause the at least one processor to determine the plurality of partial layers to remain in the gene pool in the evaluation up to i-th time, and determine the plurality of partial layers to remain in the gene pool based on evaluation values in the evaluation in (i+1)th and subsequent times, where i is a natural number smaller than n.
4. The information processing device according to claim 1 , wherein
each partial layer includes convolution.
5. The information processing device according to claim 1 , wherein
the evaluation of each partial layer includes evaluating the deep learning model into which the partial layer is incorporated for each epoch of a series of epochs, and ending learning of the deep learning model when an evaluation value becomes lower than in a previous epoch in the series of epochs.
6. The information processing device according to claim 1 , wherein
in the evaluation, the program code is further configured to cause the at least one processor to determine the plurality of partial layers to remain in the gene pool in such a way that the number of partial layers in the gene pool does not change.
7. The information processing device according to claim 1 , wherein
the genetic programming generates a new partial layer by increasing the probability of mutation applied to a partial layer having an evaluation value lower than a specified threshold.
8. The information processing device according to claim 1 , wherein
when an evaluation value of the deep learning model does not become higher, the program code is further configured to cause the at least one processor to increase a number of repetitions of a search including the generation and the evaluation, increase a number of partial layers in the gene pool, increase a probability of crossover in the genetic programming, or increase a probability of mutation in the genetic programming.
9. An information processing method performed by an information processing device including a processor, the method comprising:
acquiring a plurality of partial layers as a gene pool, the partial layers being candidates for elements of a deep learning model;
generating a new partial layer from the gene pool by using genetic programming;
evaluating each partial layer by incorporating, into a template of the deep learning model, each partial layer in the gene pool to which the new partial layer is added, and determining a plurality of partial layers to remain in the gene pool; and
outputting the deep learning model into which the partial layer with a highest evaluation value is incorporated.
10. A non-transitory recording medium storing an information processing program causing a computer to:
acquire a plurality of partial layers as a gene pool, the partial layers being candidates for elements of a deep learning model;
generate a new partial layer from the gene pool by using genetic programming;
evaluate each partial layer by incorporating, into a template of the deep learning model, each partial layer in the gene pool to which the new partial layer is added, and determine a plurality of partial layers to remain in the gene pool; and
output the deep learning model into which the partial layer with a highest evaluation value is incorporated.