Providing apparatus, providing method, and computer program product
A providing apparatus according to an embodiment of the present disclosure includes a memory and a hardware processor coupled to the memory. The hardware processor is configured to: store, in the memory, a first machine learning model capable of changing an amount of calculation of a model of a neural network; acquire device information; set, based on the device information, extraction conditions representing conditions for extracting second machine learning models from the first machine learning model; extract the second machine learning models from the first machine learning model based on the extraction conditions; and provide the second machine learning models to a device specified by the device information.
1. A providing apparatus comprising:
a memory; and
a hardware processor coupled to the memory and configured to:
store, in the memory, a first machine learning model capable of changing an amount of calculation of a model of a neural network;
acquire device information;
set, based on the device information, extraction conditions representing conditions for extracting second machine learning models from the first machine learning model;
extract the second machine learning models from the first machine learning model based on the extraction conditions; and
provide the second machine learning models to a device specified by the device information.
2. The apparatus according to claim 1 , wherein a size of each of the second machine learning models is smaller than a size of the first machine learning model.
3. The apparatus according to claim 1 , wherein the hardware processor is further configured to store, as management information in the memory, the device information and the extraction conditions in a manner of making connections with each other.
4. The apparatus according to claim 3 , wherein the hardware processor is further configured to:
train the first machine learning model; and
store, in the memory, learning information on the first machine learning model in a manner of making connections with the management information.
5. The apparatus according to claim 4 , wherein the learning information includes:
identification information for identifying the first machine learning model;
a date when the first machine learning model was generated; and
identification information for identifying a learning dataset that was used for learning of the first machine learning model.
6. The apparatus according to claim 3 , further comprising a user interface (UT) configured to receive a disclosure request for the management information and return a response satisfying a search condition specified by the disclosure request.
7. The apparatus according to claim 1 , wherein the device information includes identification information for identifying the device and specification information representing hardware specifications of the device.
8. The apparatus according to claim 7 , wherein the device information further includes control information on inference processing using the second machine learning models.
9. The apparatus according to claim 8 , wherein the control information includes at least one of:
a target amount of calculation of the inference processing executed on a device that is provided with the second machine learning models;
a target model size of the second machine learning models used for the inference processing executed on the device;
a target speed of the inference processing executed on the device; and
a target recognition rate of the inference processing executed on the device.
10. The apparatus according to claim 1 , wherein
the extraction conditions include a rank for controlling the amount of calculation of the second machine learning models, and
the hardware processor is further configured to extract the second machine learning models from the first machine learning model by:
decomposing at least one of weight matrices included in the first machine learning model into two or more matrices by using a singular value decomposition technique; and
changing a size of each of the decomposed matrices in accordance with the rank.
11. The apparatus according to claim 1 , wherein
the extraction conditions include a number of layers of the second machine learning models,
the first machine learning model includes a Residual Network (ResNet) block, and
the hardware processor is further configured to extract the second machine learning models from the first machine learning model by decomposing the ResNet block into a network representation having the number of layers specified by the extraction conditions while treating the ResNet block as ordinary differential equations.
12. A providing method implemented by a computer, the method comprising:
reading out, from a memory, a first machine learning model capable of changing an amount of calculation of a model of a neural network;
acquiring device information;
setting, based on the device information, extraction conditions representing conditions for extracting second machine learning models from the first machine learning model;
extracting the second machine learning models from the first machine learning model based on the extraction conditions; and
providing the second machine learning models to a device specified by the device information.
13. A computer program product comprising a non-transitory computer-readable recording medium on which an executable program is recorded, the program instructing a computer to:
store, in a memory, a first machine learning model capable of changing an amount of calculation of a model of a neural network;
acquire device information;
set, based on the device information, extraction conditions representing conditions for extracting second machine learning models from the first machine learning model;
extract the second machine learning models from the first machine learning model based on the extraction conditions; and
provide the second machine learning models to a device specified by the device information.