# US 2019/0377772 A1 · Systems and methods for converting discrete wavelets to tensor fields and using neural networks to process tensor fields

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

- Number: US 2019/0377772 A1
- Document ID: US20190377772A1
- Status: Published application
- Kind: A1
- Application number: 16/549,557
- Filed: 2019-08-23
- Published: 2019-12-12
- Earliest claimed: 2018-02-08
- Assignees: Deep Labs Inc.
- Applicants: Deep Labs Inc.
- Inventors: Patrick Faith
- Web page: https://patentgrep.com/patent/US20190377772A1

## CPC

- G06F 17/148: {Wavelet transforms}
- G06N 3/02: Neural networks
- G06N 3/045: Combinations of networks
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The present disclosure relates to systems and methods for detecting and identifying anomalies within a discrete wavelet database. In one implementation, the system may include one or more memories storing instructions and one or more processors configured to execute the instructions. The instructions may include instructions to receive a new wavelet, convert the net transaction to a wavelet, convert the wavelet to a tensor using an exponential smoothing average, calculate a difference field between the tensor and a field having one or more previous transactions represented as tensors, perform a weighted summation of the difference field to produce a difference vector, apply one or more models to the difference vector to determine a likelihood of the new wavelet representing an anomaly, and add the new wavelet to the field when the likelihood is below a threshold.

## Claims

### Claim at position 1

-6. (canceled)

### Claim 7

A system for training a deep field network to detect anomalies within a database comprising discrete wavelets, the system comprising:
one or more memories storing instructions; and
one or more processors configured to execute the instructions to:
receive a plurality of transactions,
convert each transaction to a corresponding wavelet,
group the plurality of wavelets and corresponding tensors by coefficients included in the wavelets,
train a neural network for each group independently of other groups, and
integrate the neural networks into a deep field network.

### Claim 8 (depends on 7)

The system of claim 7, wherein each neural network is trained specific to at least one of a particular account, a particular location, and a particular time of day.

### Claim 9 (depends on 7)

The system of claim 7, wherein at least one wavelet is grouped in at least two groups.

### Claim 10 (depends on 7)

The system of claim 7, wherein at least one of the transactions comprises at least one of a purchase transaction, an account enrollment, and a portfolio rebalancing.

### Claim 11 (depends on 7)

The system of claim 7, wherein the corresponding tensors are obtained using a discrete wavelet transform.

### Claim 12 (depends on 11)

The system of claim 11, wherein the discrete wavelet transform is performed using a filter bank having a plurality of convolutional-accumulators.

### Claim 13 (depends on 12)

The system of claim 12, wherein the convolutional accumulators are configured to accumulate using base e.

### Claim 14 (depends on 12)

The system of claim 12, wherein the discrete wavelet transform includes an exponential smoothing average in the filter bank.

### Claim 15

A system for authorizing a transaction using cascading discrete wavelets, the system comprising:
one or more memories storing instructions; and
one or more processors configured to execute the instructions to:
receive a new transaction,
convert the new transaction to a wavelet,
convert the wavelet to a tensor using an exponential smoothing average,
calculate a difference field between the tensor and a field having one or more previous transactions represented as tensors,
perform a weighted summation of the difference field to produce a difference vector,
apply one or more models to the difference vector to determine a likelihood of the transaction being high risk, and
authorize the new transaction when the likelihood is below a threshold.

### Claim 16 (depends on 15)

The system of claim 15, wherein the threshold is predetermined.

### Claim 17 (depends on 15)

The system of claim 15, wherein the threshold is dynamically determined based on the one or more models.

### Claim 18 (depends on 15)

The system of claim 15, wherein the one or more models are selected based on coefficients included in the tensor.

### Claim 19 (depends on 15)

The system of claim 15, wherein the likelihood is determined based on a magnitude of the difference vector and a direction of the difference vector.

### Claim 20 (depends on 15)

The system of claim 15, wherein converting the wavelet to a tensor includes applying a discrete wavelet transform.

### Claim 21 (depends on 20)

The system of claim 20, wherein the discrete wavelet transform includes a filter bank comprising a plurality of convolutional-accumulators.

### Claim 22 (depends on 21)

The system of claim 21, wherein the convolutional accumulators are configured to accumulate using base e.

### Claim 23 (depends on 21)

The system of claim 21, wherein the discrete wavelet transform includes the exponential smoothing average in the filter bank.

### Claim 24 (depends on 7)

The system of claim 7, wherein the plurality of transactions are encoded in one or more data serialization formats.

### Claim 25 (depends on 24)

The system of claim 24, wherein converting each transaction to a corresponding wavelet comprises converting each transaction into at least one of the one or more data serialization formats.

### Claim 26 (depends on 7)

The system of claim 7, wherein integrating the neural networks comprises combining the neural networks such that, when a tensor is applied to the neural networks, one of the neural networks trained on a group having more similar coefficients is amplified while other of the neural networks trained on a group having less similar coefficients is minimized.
