# US 2021/0174167 A1 · Custom convolutional neural network architectures for exposure detection

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

- Number: US 2021/0174167 A1
- Document ID: US20210174167A1
- Status: Published application
- Kind: A1
- Application number: 16/707,916
- Filed: 2019-12-09
- Published: 2021-06-10
- Earliest claimed: 2019-12-09
- Assignees: Bank of America Corporation
- Applicants: Bank of America Corporation
- Inventors: Eren Kursun; Hongda Shen
- Web page: https://patentgrep.com/patent/US20210174167A1

## CPC

- G06N 3/045: Combinations of networks
- G06N 3/086: using evolutionary algorithms, e.g. genetic algorithms or genetic programming
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections
- G10L 15/16: using artificial neural networks
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A system is typically configured for customizing interconnectivity of one or more layers associated with a neural network architecture, wherein the neural network architecture is associated with an application, customizing functional transformation of the one or more layers associated with the neural network architecture, wherein each of the one or more layers comprises a custom transformation function, and generating a custom neural network architecture based on customizing the interconnectivity and the functional transformation of the one or more layers.

## Claims

### Claim 1

A system for generating custom neural network architectures, comprising:
a multi-layer device comprising one or more neural networks, wherein the system is configured to alter the one or more neural networks by:
customizing interconnectivity of one or more layers associated with a neural network architecture, wherein the neural network architecture is associated with an application;
customizing functional transformation of the one or more layers associated with the neural network architecture, wherein each of the one or more layers comprises a custom transformation function; and
generating a custom neural network architecture based on customizing the interconnectivity and the functional transformation of the one or more layers.

### Claim 2 (depends on 1)

The system of claim 1, wherein customizing the functional transformation of the one or more layers comprises dynamically altering the custom transformation function associated with each of the one or more layers, wherein the custom transformation function associated with each of the one or more layers is a function of time.

### Claim 3 (depends on 1)

The system of claim 1, wherein customization of the interconnectivity and the functional transformation of the one or more layers is based on data associated with the application and a type of the application.

### Claim 4 (depends on 1)

The system of claim 1, wherein the neural network architecture is a convolutional neural network architecture.

### Claim 5 (depends on 1)

The system of claim 1, wherein the neural network architecture is a non-convolutional neural network architecture.

### Claim 6 (depends on 1)

The system of claim 1, wherein the system is further configured to optimize the custom neural network architecture based on one or more metrics and an optimization algorithm.

### Claim 7 (depends on 6)

The system of claim 6, wherein optimizing the custom neural network comprises performing incremental updates on the custom neural network architecture.

### Claim 8 (depends on 1)

The system of claim 1, wherein the multi-layer device is a multi-layer silicon chip.

### Claim 9

A computer program product for generating custom neural network architectures, the computer program product comprising instructions for:
customizing interconnectivity of one or more layers associated with a neural network architecture, wherein the neural network architecture is associated with an application;
customizing functional transformation of the one or more layers associated with the neural network architecture, wherein each of the one or more layers comprises a custom transformation function; and
generating a custom neural network architecture based on customizing the interconnectivity and the functional transformation of the one or more layers.

### Claim 10 (depends on 9)

The computer program product of claim 9, wherein customization of the interconnectivity and the functional transformation of the one or more layers is based on data associated with the application and a type of the application.

### Claim 11 (depends on 9)

The computer program product of claim 9, wherein the computer-executable instructions are further configured for optimizing the custom neural network architecture is based on one or more metrics and an optimization algorithm.

### Claim 12 (depends on 9)

The computer program product of claim 9, wherein the neural network architecture is a non-convolutional neural network architecture.

### Claim 13 (depends on 9)

The computer program product of claim 9, wherein the neural network architecture is a convolutional neural network architecture.

### Claim 14 (depends on 9)

The computer program product of claim 9, wherein the customizing the interconnectivity of the one or more layers allows neurons in a first layer of the one or more layers to communicate with all the one or more layers.

### Claim 15

A computer implemented method for generating custom neural network architectures, wherein the method comprises:
customizing interconnectivity of one or more layers associated with a neural network architecture, wherein the neural network architecture is associated with an application;
customizing functional transformation of the one or more layers associated with the neural network architecture, wherein each of the one or more layers comprises a custom transformation function; and
generating a custom neural network architecture based on customizing the interconnectivity and the functional transformation of the one or more layers.

### Claim 16 (depends on 15)

The computer implemented method of claim 15, wherein customization of the interconnectivity and the functional transformation of the one or more layers is based on data associated with the application and a type of the application.

### Claim 17 (depends on 15)

The computer implemented method of claim 15, wherein optimizing the custom neural network architecture is based on one or more metrics and an optimization algorithm.

### Claim 18 (depends on 15)

The computer implemented method of claim 15, wherein the neural network architecture is a convolutional neural network architecture.

### Claim 19 (depends on 15)

The computer implemented method of claim 15, wherein the neural network architecture is a non-convolutional neural network architecture.

### Claim 20 (depends on 15)

The computer implemented method of claim 15, wherein the customizing the interconnectivity of the one or more layers allows neurons in a first layer of the one or more layers to communicate with all the one or more layers.
