# US 2026/0087304 A1 · Neural Networks with Semantic Inference

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

- Number: US 2026/0087304 A1
- Document ID: US20260087304A1
- Status: Published application
- Kind: A1
- Application number: 18/893,610
- Filed: 2024-09-23
- Published: 2026-03-26
- Earliest claimed: 2023-09-22
- Assignees: Northeastern University; Government of the United States as Represented by the Secretary of the Air Force
- Applicants: Northeastern University; Government of the United States as represented by the Secretary of the Air Force
- Inventors: Francesco Restuccia; Jonathan Ashdown; A Q M Sazzad Sayyed
- Web page: https://patentgrep.com/patent/US20260087304A1

## CPC

- G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks
- G06F 18/2415: based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/09: Supervised learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Systems and method of classification are provided. Upon receiving an input, a feature set is defined from the input. A semantic cluster to be associated with the input is defined based on the feature set, the semantic cluster being one of a plurality of semantic clusters each defining a subset of outputs of a neural network based on semantic similarity of the subset. The feature set is applied to a subgraph corresponding to the semantic cluster, the subgraph being one of a plurality of subgraphs each defining a portion of the neural network. A classification for the input is then be determined based on an output of the subgraph.

## Claims

### Claim 1

A method of classification, comprising:
defining a feature set from an input;
identifying a semantic cluster to be associated with the input based on the feature set, the semantic cluster being one of a plurality of semantic clusters each defining a subset of outputs of a neural network based on semantic similarity of the subset;
applying the feature set to a subgraph corresponding to the semantic cluster, the subgraph being one of a plurality of subgraphs each defining a portion of the neural network; and
determining a classification for the input based on an output of the subgraph.

### Claim 2 (depends on 1)

The method of claim 1, wherein nodes of the subgraph correspond to a subset of filters of the neural network, the subset of filters having an activation rate that exceeds a threshold value for members of the semantic cluster.

### Claim 3 (depends on 1)

The method of claim 1, wherein nodes of the subgraph correspond to filters of the neural network, and wherein the subgraph includes at least one node common to another one of the plurality of subgraphs.

### Claim 4 (depends on 1)

The method of claim 1, wherein the semantic cluster is identified via at least one layer of the neural network.

### Claim 5 (depends on 1)

The method of claim 1, wherein the subgraph encompasses a set of filters of the neural network that are excluded by at least one of the plurality of subgraphs.

### Claim 6 (depends on 1)

The method of claim 1, further comprising storing a subset of the subgraphs to a memory device independent of a remainder of the subgraphs.

### Claim 7

A system for classification, comprising:
a feature extractor configured to define a feature set from an input;
a predictor configured to identify a semantic cluster to be associated with the input based on the feature set, the semantic cluster being one of a plurality of semantic clusters each defining a subset of outputs of a neural network based on semantic similarity of the subset;
a plurality of subgraphs each defining a portion of the neural network; and
a router configured to apply the feature set to a subgraph corresponding to the semantic cluster, the subgraph being one of a plurality of the plurality of subgraphs;
the subgraph being configured to generate an output for determining a classification for the input.

### Claim 8 (depends on 7)

The system of claim 7, wherein nodes of the subgraph correspond to a subset of filters of the neural network, the subset of filters having an activation rate that exceeds a threshold value for members of the semantic cluster.

### Claim 9 (depends on 7)

The system of claim 7, wherein nodes of the subgraph correspond to filters of the neural network, and wherein the subgraph includes at least one node common to another one of the plurality of subgraphs.

### Claim 10 (depends on 7)

The system of claim 7, wherein the semantic cluster is identified via at least one layer of the neural network.

### Claim 11 (depends on 7)

The system of claim 7, wherein the subgraph encompasses a set of filters of the neural network that are excluded by at least one of the plurality of subgraphs.

### Claim 12 (depends on 7)

The system of claim 7, wherein a subset of the subgraphs are stored to a memory device independent of a remainder of the subgraphs.

### Claim 13

A method of optimizing a neural network, comprising:
defining a plurality of semantic clusters for a dataset of inputs to the neural network, each of the semantic clusters including a subset of the inputs based on semantic similarity of the subset;
defining a plurality of subgraphs each corresponding to one of the semantic clusters, each of the subgraphs being a portion of the neural network; and
generating a router configured to 1) associate an input with one of the semantic clusters, and 2) apply the input to the associated semantic cluster.

### Claim 14 (depends on 13)

The method of claim 13, wherein nodes of the subgraphs correspond to a subset of filters of the neural network, the subset of filters having an activation rate that exceeds a threshold value for members of the semantic cluster.

### Claim 15 (depends on 13)

The method of claim 13, wherein nodes of the subgraph correspond to filters of the neural network, and wherein the subgraph includes at least one node common to another one of the plurality of subgraphs.

### Claim 16 (depends on 13)

The method of claim 13, wherein the associated semantic cluster is identified via at least one layer of the neural network.

### Claim 17 (depends on 13)

The method of claim 13, wherein each of the plurality of subgraph encompasses a set of filters of the neural network that are excluded by at least one of the plurality of subgraphs.

### Claim 18 (depends on 13)

The method of claim 13, further comprising storing a subset of the subgraphs to a memory device independent of a remainder of the subgraphs.
