# US 2023/0274138 A1 · Data Object Classification Using an Optimized Neural Network

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2023/0274138 A1
- Document ID: US20230274138A1
- Status: Published application
- Kind: A1
- Application number: 18/143,297
- Filed: 2023-05-04
- Published: 2023-08-31
- Earliest claimed: 2019-11-15
- Assignees: Disney Enterprises, Inc.; ETH Zürich (Eidgenössische Technische Hochschule Zürich)
- Applicants: Disney Enterprises, Inc.; ETH Zürich (EIDGENÖSSISCHE TECHNISCHE HOCHSCHULE ZÜRICH)
- Inventors: Hayko Jochen Wilhelm Riemenschneider; Leonhard Markus Helminger; Christopher Richard Schroers; Abdelaziz Djelouah
- Web page: https://patentgrep.com/patent/US20230274138A1

## CPC

- G06N 3/08: Learning methods
- G06F 18/2155: {characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling}
- G06F 18/24: Classification techniques
- G06N 3/045: Combinations of networks
- G06N 3/048: Activation functions
- G06N 3/0499: Feedforward networks
- G06N 3/09: Supervised learning
- G06V 10/75: Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V 10/764: using classification, e.g. of video objects
- G06V 10/7753: {Incorporation of unlabelled data, e.g. multiple instance learning [MIL]}
- G06V 10/776: Validation; Performance evaluation
- G06V 10/82: using neural networks
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A system includes a computing platform having a hardware processor and a memory storing a software code and a neural network (NN) having multiple layers including a last activation layer and a loss layer. The hardware processor executes the software code to identify different combinations of layers for testing the NN, each combination including candidate function(s) for the last activation layer and candidate function(s) for the loss layer. For each different combination, the software code configures the NN based on the combination, inputs, into the configured NN, a training dataset including multiple data objects, receives, from the configured NN, a classification of the data objects, and generates a performance assessment for the combination based on the classification. The software code determines a preferred combination of layers for the NN including selected candidate functions for the last activation layer and the loss layer, based on a comparison of the performance assessments.

## Claims

### Claim at position 1

1-20. (canceled)

### Claim 21

A system comprising:
a hardware processor; and
a system memory storing a software code;
the hardware processor configured to execute the software code to:
obtain a plurality of real images;
composite the plurality of real images to form a montage of the plurality of real images;
identify a plurality of labels for association with the montage; and
label the montage using one or more of the plurality of identified labels to generate the plurality of images in a training dataset, wherein noise is parametrically introduced into the training dataset, resulting in a subset of the plurality of images being purposely mislabeled.

### Claim 22 (depends on 21)

The system of claim 21, wherein a plurality of parameters utilized to introduce the noise into the training dataset comprise a number of peaks (tip) in the training dataset.

### Claim 23 (depends on 21)

The system of claim 21, wherein a plurality of parameters utilized to introduce the noise into the training dataset comprise a likelihood of noise (pn) in the training dataset.

### Claim 24 (depends on 21)

The system of claim 21, wherein a plurality of parameters utilized to introduce the noise into the training dataset comprise a balance (pa) between false positives and false negatives in the training dataset.

### Claim 25 (depends on 21)

The system of claim 21, wherein the plurality of real images are obtained from the Modified National Institute of Science and Technology (MNIST) dataset.

### Claim 26 (depends on 21)

The system of claim 21, the hardware processor is further configured to execute the software code to:
configure the NN using a combination of an activation candidate function and a first loss candidate function;
input the training dataset into the NN configured using the combination; and
receive, from the NN, a classification of the training dataset.

### Claim 27 (depends on 26)

The system of claim 26, the hardware processor is further configured to execute the software code to:
determine, based on the classification, a preference for the combination of the activation candidate function and the first loss candidate function.

### Claim 28 (depends on 26)

The system of claim 26, wherein the activation candidate function is a softmax activation function, and wherein the loss candidate function follows the softmax activation function.

### Claim 29 (depends on 26)

The system of claim 26, wherein the combination further includes a normalization candidate function.

### Claim 30 (depends on 26)

The system of claim 26, wherein the activation candidate function is one of a sigmoid activation function or an L1-normalization function.

### Claim 31

A computer-implemented method comprising:
obtaining a plurality of real images;
compositing the plurality of real images to form a montage of the plurality of real images;
identifying a plurality of labels for association with the montage; and
labeling the montage using one or more of the plurality of identified labels to generate the plurality of images in a training dataset, wherein noise is parametrically introduced into the training dataset, resulting in a subset of the plurality of images being purposely mislabeled.

### Claim 32 (depends on 31)

The computer-implemented of claim 31, wherein a plurality of parameters utilized to introduce the noise into the training dataset comprise a number of peaks (dp) in the training dataset.

### Claim 33 (depends on 31)

The computer-implemented of claim 31, wherein a plurality of parameters utilized to introduce the noise into the training dataset comprise a likelihood of noise (pn) in the training dataset.

### Claim 34 (depends on 31)

The computer-implemented of claim 31, wherein a plurality of parameters utilized to introduce the noise into the training dataset comprise a balance (pa) between false positives and false negatives in the training dataset.

### Claim 35 (depends on 31)

The computer-implemented of claim 31, wherein the plurality of real images are obtained from the Modified National Institute of Science and Technology (MNIST) dataset.

### Claim 36 (depends on 31)

The computer-implemented of claim 31, further comprising:
configuring the NN using a combination of an activation candidate function and a first loss candidate function;
inputting the training dataset into the NN configured using the combination; and
receiving, from the NN, a classification of the training dataset.

### Claim 37 (depends on 36)

The computer-implemented of claim 36, further comprising:
determining, based on the classification, a preference for the combination of the activation candidate function and the first loss candidate function.

### Claim 38 (depends on 36)

The computer-implemented of claim 36, wherein the activation candidate function is a softmax activation function, and wherein the loss candidate function follows the softmax activation function.

### Claim 39 (depends on 36)

The computer-implemented of claim 36, wherein the combination further includes a normalization candidate function.

### Claim 40 (depends on 36)

The computer-implemented of claim 36, wherein the activation candidate function is one of a sigmoid activation function or an L1-normalization function.
