# US 2022/0277188 A1 · Systems and methods for classifying data sets using corresponding neural networks

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

- Number: US 2022/0277188 A1
- Document ID: US20220277188A1
- Status: Published application
- Kind: A1
- Application number: 17/663,529
- Filed: 2022-05-16
- Published: 2022-09-01
- Earliest claimed: 2019-08-06
- Assignees: Capital One Services, LLC
- Applicants: Capital One Services, LLC
- Inventors: Austin WALTERS; Jeremy GOODSITT
- Web page: https://patentgrep.com/patent/US20220277188A1

## CPC

- G06N 3/08: Learning methods
- G06F 18/24: Classification techniques
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0475: Generative networks
- G06N 3/09: Supervised learning
- G06N 3/094: Adversarial learning
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V 10/82: using neural networks
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/048: Activation functions
- 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

The present disclosure relates to systems and methods for classifying data sets using associated functions from neural networks. In one example, a system for classifying data sets by corresponding functions includes at least one processor and at least one non-transitory memory storing instructions that, when executed by the at least one processor cause the system to perform operations including: obtaining a neural network associated with a data set, the neural network being trained to generate synthetic data sets related to the data set; selecting a test set of inputs to the neural network; obtaining a corresponding set of outputs by applying the neural network to the test set of inputs; estimating one or more functions describing the test set of inputs and the corresponding set of outputs; and indexing the estimated one or more functions to the data.

## Claims

### Claim at position 1

1-20. (canceled)

### Claim 21

A system for classifying data sets by corresponding functions, comprising:
at least one processor; and
at least one non-transitory memory storing instructions that, when executed by the at least one processor cause the system to perform operations comprising:
obtaining a neural network trained to generate synthetic data sets having at least one characteristic of a first data set;
selecting a test set of inputs for the neural network;
obtaining, by applying the neural network to the test set of inputs, a corresponding set of outputs from the neural network the neural network outputs comprising a second data set, the second data set comprising synthetic data;
estimating at least one function associated with one or more features of the test set of inputs more likely to have resulted in the corresponding set of neural network outputs than other features of the test set of inputs; and
indexing the estimated at least one function to the first data set.

### Claim 22 (depends on 21)

The system of claim 21, wherein estimating the at least one function comprises assessing commonality of features across the test set of inputs and the corresponding set of outputs and identifying one or more features in the test set of inputs that are more likely to be explanatory of the corresponding set of outputs.

### Claim 23 (depends on 21)

The system of claim 21 wherein estimating the at least one function comprises comparing a distance between the test set of inputs and a calculated function.

### Claim 24 (depends on 21)

The system of claim 21 wherein estimating the at least one function comprises analyzing a commonality between the test set of inputs and the corresponding set of outputs.

### Claim 25 (depends on 21)

The system of claim 21, wherein the operations further comprise storing the indexed at least one function and data sets in a function index.

### Claim 26 (depends on 25)

The system of claim 25, wherein the operations further comprise generating a relational mapping for searching a received data set based on the estimated at least one function.

### Claim 27 (depends on 25)

The system of claim 25, wherein the operations further comprise:
receiving, from an interface, a search query, the search query comprising an example data set, one or more statistical measures, or other structural descriptions of a desired data set;
searching the function index based on the search query; and
returning, based on the search, indexed data sets matching the search query.

### Claim 28

A computer-implemented method for classifying data sets by corresponding functions comprising:
obtaining a neural network trained to generate synthetic data sets having at least one characteristic of a first data set;
selecting a test set of inputs for the neural network;
obtaining, by applying the neural network to the test set of inputs, a corresponding set of outputs, from the neural network the neural network outputs comprising a second data set, the second data set comprising synthetic data;
estimating at least one function associated with one or more features of the test set of inputs more likely to have resulted in the corresponding set of neural network outputs than other features of the test set of inputs; and
indexing the estimated at least one function to the first data set.

### Claim 29 (depends on 28)

The method of claim 28 wherein estimating the at least one function comprises comparing a distance between the test set of inputs and a calculated function.

### Claim 30 (depends on 29)

The method of claim 29 wherein estimating the at least one function further comprises utilizing a greedy algorithm to extract the one or more features in a particular order.

### Claim 31 (depends on 28)

The method of claim 28 wherein indexing the estimated at least one function further comprises generating a relational index.

### Claim 32 (depends on 28)

The method of claim 28 wherein indexing the estimated at least one function further comprises generating a graphical index.

### Claim 33 (depends on 28)

The method of claim 28 wherein the operations further comprise clustering the indexed data sets using the at least one function, the clustering comprising at least one of a hierarchical clustering, a centroid-based clustering, a distribution-based clustering, or a density-based clustering.

### Claim 34 (depends on 28)

The method of claim 28, wherein the neural network comprises at least one of a convolutional neural network, a recurrent neural network, an auto-encoder, a variational auto-encoder, or a generative adversarial network.

### Claim 35

A system for classifying data sets by corresponding functions, comprising:
at least one processor; and
at least one non-transitory memory storing instructions that, when executed by the at least one processor cause the system to perform operations comprising:
obtaining a first data set;
training, using the first data set, a neural network to generate synthetic data sets having a characteristic associated with the first data set;
selecting a test set of inputs for the neural network;
obtaining, by applying the neural network to the test set of inputs, a corresponding set of outputs, from the neural network the neural network outputs comprising a second data set, the second data set comprising synthetic data;
estimating at least one function associated with one or more features of the test set of inputs more likely to have resulted in the corresponding set of neural network outputs than other features of the test set of inputs;
indexing the estimated at least one function to the first data set; and
storing the indexed at least one function and data sets in a function index.

### Claim 36 (depends on 35)

The system of claim 35, wherein training the neural network comprises adjusting activation functions of the neural network to reduce an associated loss function.

### Claim 37 (depends on 35)

The system of claim 35, wherein the operations further comprise:
receiving a search query indicating at least one search query function;
searching the function index based on the search query; and
returning, based on the search, the first data set.

### Claim 38 (depends on 35)

The system of claim 35, wherein the operations further comprise:
identifying one or more activation functions within a node of a branch of a neural network triggered by the one or more features; and
mapping the one or more features to a branch of the neural network.

### Claim 39 (depends on 35)

The system of claim 35, wherein searching the function index based on the search query comprises:
clustering one or more functions of the function index to form a plurality of function clusters; and
matching the search query function to a first cluster of the plurality of clusters.

### Claim 40 (depends on 39)

The system of claim 39, wherein matching the search query function to a first cluster comprises:
generating a score for the first cluster indicating a degree of belongingness of the search query function to the first cluster; and
determining that the score exceeds a predetermined threshold.
