Learning method, and learning apparatus, and recording medium

Inventors: Takashi Katoh (Kawasaki, JP); Kento Uemura (Kawasaki, JP); Suguru Yasutomi (Kawasaki, JP); Takeshi Osoekawa (Ohta, JP)

Assignee: FUJITSU LIMITED

G06N3/084 G06N3/0454 G06N20/20

View Patent ↗

Loading inventors, assignments & file history…

Monitor This Case

Get email alerts when status or documents change.

Order Certified Copies

Most orders are placed with the USPTO same day — all within 24 business hours.

Order via The Patent Place →

Pre-filled with this patent's details

Quick Facts

Patent No.: US 11,620,530
App. No.: 16/741,860
Granted: Apr 4, 2023
Kind: B2

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

A learning method executed by a computer, the learning method includes: learning parameters of a machine learning model having intermediate feature values by inputting a plurality of augmented training data, which is generated by augmenting original training data, to the machine learning model so that specific intermediate feature values, which are calculated from specific augmented training data augmented from a same original training data, become similar to each other.

Claims (25)

1. A learning method executed by a computer, the learning method comprising:

obtaining a first training data set and a second training data set, the first training data set including a plurality of first augmented training data each of which is generated by augmenting first original training data, the second training data set including a plurality of second augmented training data each of which is generated by augmenting second original training data; and

training parameters of a machine learning model by inputting each of the first training data set and the second training data set, to the machine learning model, the training of the parameters including

in response to the inputting of the first training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model so that first intermediate feature values, which are calculated by the machine learning model from the first training data set, become similar to each other, and

in response to the inputting of the second training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model so that second intermediate feature values, which are calculated by the machine learning model from the second training data set, become similar to each other and become different from each of the first intermediate feature values.

2. The learning method according to claim 1 , wherein the training of the parameters of the machine learning model includes:

in response to the inputting of each of the first training data set and the second training data set to the machine learning model, causing the machine learning model to calculate variance of the intermediate feature values output from a lower layer of the machine learning model, and to train the parameters of the machine learning model to reduce the variance.

3. The learning method according to claim 1 , wherein the training of the parameters of the machine learning model includes:

in response to the inputting of the first training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model and a first reference feature value, the first reference feature value corresponding to the first original training data, so that the first intermediate feature values output from a lower layer of the machine learning model and the first reference feature value become similar to each other; and

in response to the inputting of the second training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model and a second reference feature value, the second reference feature value corresponding to the second original training data, so that the second intermediate feature values output from a lower layer of the machine learning model and the second reference feature value become similar to each other.

4. The learning method according to claim 1 , wherein the training of the parameters of the machine learning model includes:

in response to the inputting of the first training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model and a first reference feature value, the first reference feature value corresponding to the first original training data, so that a distribution of the first intermediate feature values output from a lower layer of the machine learning model and the first reference feature value become similar to each other; and

in response to the inputting of the second training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model and a second reference feature value, the second reference feature value corresponding to the second original training data, so that a distribution of the second intermediate feature values output from a lower layer of the machine learning model and the second reference feature value become similar to each other.

5. A learning apparatus comprising:

a memory; and

a processor coupled to the memory, the processor configured to:

obtain a first training data set and a second training data set, the first training data set including a plurality of first augmented training data each of which is generated by augmenting first original training data, the second training data set including a plurality of second augmented training data each of which is generated by augmenting second original training data; and

train parameters of a machine learning model by inputting each of the first training data set and the second training data set to the machine learning model, the training of the parameters including

in response to the inputting of the first training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model so that first intermediate feature values, which are calculated by the machine learning model from the first training data set, become similar to each other, and

6. A non-transitory computer-readable recording medium having stored therein a program that causes a computer to execute a learning process, the learning process comprising:

training parameters of a machine learning model by inputting each of the first training data set and the second training data set to the machine learning model, the training of the parameters including

in response to the inputting of the first training data set to the machine learning model, causing the machine learning model to train the parameters of the machine learning model so that first intermediate feature values, which are calculated by the machine learning model from the first training data set, become similar to each other, and

Assignments (1)

ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jan 14, 2020

From: KATOH, TAKASHI; UEMURA, KENTO; YASUTOMI, SUGURU; OSOEKAWA, TAKESHI

To: FUJITSU LIMITED

Reel/Frame 052585/0039 →

Priority Claims (1)

JP JP2019-006422 · Jan 17, 2019 · national

Continuity (1)

Related Publication 20200234140A1 · Jul 23, 2020

Cited By (3)

US 12,400,434 US 12,524,501 US 12,554,796