# US 2025/0095281 A1 · Systems and methods for recursive dimensional decomposition convolution

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

- Number: US 2025/0095281 A1
- Document ID: US20250095281A1
- Status: Published application
- Kind: A1
- Application number: 18/467,451
- Filed: 2023-09-14
- Published: 2025-03-20
- Earliest claimed: 2023-09-14
- Assignees: QUALCOMM Incorporated
- Applicants: QUALCOMM Incorporated
- Inventors: Jamie Menjay LIN; Kai WANG
- Web page: https://patentgrep.com/patent/US20250095281A1

## CPC

- G06F 17/153: {Multidimensional correlation or convolution}
- G06T 15/20: Perspective computation
- G06N 3/00: Computing arrangements based on biological models
- G06N 3/045: Combinations of networks
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Dimensional decomposition convolution systems and techniques are described. A system receives a tensor including a first number of dimensions. The system processes a variant of the tensor using a convolution function including a second number of dimensions to generate a processed tensor. The first number of dimensions is greater than the second number of dimensions. A plurality of dimensions from the first number of dimensions of the tensor are grouped into a dimension of the variant of the tensor to reduce dimensionality of the variant of the tensor to the second number of dimensions.

## Claims

### Claim 1

An apparatus for dimensional decomposition convolution, the apparatus comprising:
at least one memory; and
at least one processor coupled to the at least one memory and configured to:
receive a tensor including a first number of dimensions; and
process a variant of the tensor using a convolution function including a second number of dimensions to generate a processed tensor, the first number of dimensions being greater than the second number of dimensions,
wherein a plurality of dimensions from the first number of dimensions of the tensor are grouped into a dimension of the variant of the tensor to reduce dimensionality of the variant of the tensor to the second number of dimensions.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
convolve the variant of the tensor using the convolution function to process the variant of the tensor using the convolution function.

### Claim 3 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
select a particular dimension of the first number of dimensions, the particular dimension distinct from the plurality of dimensions that are grouped, wherein the particular dimension represents a second dimension of the variant of the tensor.

### Claim 4 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
process a second variant of the tensor using the convolution function to generate a second processed tensor,
wherein a second plurality of dimensions from the first number of dimensions of the tensor are grouped into a dimension of the second variant of the tensor to reduce dimensionality of the second variant of the tensor to the second number of dimensions, and
wherein the second variant of the tensor includes the processed tensor representing one of the second plurality of dimensions.

### Claim 5 (depends on 4)

The apparatus of claim 4, wherein the at least one processor is configured to:
convolve the second variant of the tensor using the convolution function to process the second variant of the tensor using the convolution function.

### Claim 6 (depends on 4)

The apparatus of claim 4, wherein the at least one processor is configured to:
select a second specified dimension of the first number of dimensions, the second specified dimension distinct from the second plurality of dimensions that are grouped, wherein the second specified dimension represents a second dimension of the second variant of the tensor.

### Claim 7 (depends on 4)

The apparatus of claim 4, wherein the second plurality of dimensions include at least one batched processed dimension, the at least one batched processed dimension including at least the processed tensor.

### Claim 8 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
iteratively convolve a plurality of variations of the tensor using the convolution function according to an equation to convolve the variant of the tensor, wherein the equation is Conv(D)≙Conv(D−1)∘Conv(1), wherein D represents the first number of dimensions.

### Claim 9 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
iteratively convolve a plurality of variations of the tensor using the convolution function according to an equation to convolve the variant of the tensor, wherein the equation is Conv(D)≙Conv(1)∘Conv(D−1), wherein D represents the first number of dimensions.

### Claim 10 (depends on 1)

The apparatus of claim 1,
wherein the second number of dimensions is 2, and
wherein the first number of dimensions is at least 3.

### Claim 11 (depends on 1)

The apparatus of claim 1,
wherein the second number of dimensions is 3, and
wherein the first number of dimensions is at least 4.

### Claim 12 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
receive image data, wherein the tensor is associated with the image data; and
process the image data to generate processed image data, wherein the processed tensor is associated with the processed image data.

### Claim 13 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
receive video data, wherein the tensor is associated with the video data; and
process the video data to generate processed video data, wherein the processed tensor is associated with the processed video data.

### Claim 14 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
receive audio data, wherein the tensor is associated with the audio data; and
process the audio data to generate processed audio data, wherein the processed tensor is associated with the processed audio data.

### Claim 15 (depends on 1)

The apparatus of claim 1, wherein the at least one processor is configured to:
receive text data, wherein the tensor is associated with the text data; and
process the text data to generate processed text data, wherein the processed tensor is associated with the processed text data.

### Claim 16 (depends on 1)

The apparatus of claim 1, wherein the processed tensor is associated with Neural Radiance Fields (NeRF).

### Claim 17

A method of dimensional decomposition convolution, the method comprising:
receiving a tensor including a first number of dimensions; and
processing a variant of the tensor using a convolution function including a second number of dimensions to generate a processed tensor, the first number of dimensions being greater than the second number of dimensions,
wherein a plurality of dimensions from the first number of dimensions of the tensor are grouped into a dimension of the variant of the tensor to reduce dimensionality of the variant of the tensor to the second number of dimensions.

### Claim 18 (depends on 17)

The method of claim 17, further comprising:
convolving the variant of the tensor using the convolution function to process the variant of the tensor using the convolution function.

### Claim 19 (depends on 17)

The method of claim 17, further comprising:
selecting a particular dimension of the first number of dimensions, the particular dimension distinct from the plurality of dimensions that are grouped, wherein the particular dimension represents a second dimension of the variant of the tensor.

### Claim 20 (depends on 17)

The method of claim 17, further comprising:
processing a second variant of the tensor using the convolution function to generate a second processed tensor,
wherein a second plurality of dimensions from the first number of dimensions of the tensor are grouped into a dimension of the second variant of the tensor to reduce dimensionality of the second variant of the tensor to the second number of dimensions, and
wherein the second variant of the tensor includes the processed tensor representing one of the second plurality of dimensions.

### Claim 21 (depends on 20)

The method of claim 20, further comprising:
selecting a second specified dimension of the first number of dimensions, the second specified dimension distinct from the second plurality of dimensions that are grouped, wherein the second specified dimension represents a second dimension of the second variant of the tensor.

### Claim 22 (depends on 20)

The method of claim 20, wherein the second plurality of dimensions include at least one batched processed dimension, the at least one batched processed dimension including at least the processed tensor.

### Claim 23 (depends on 17)

The method of claim 17, further comprising:
iteratively convolving a plurality of variations of the tensor using the convolution function according to an equation to convolve the variant of the tensor, wherein the equation is Conv(D)≙Conv(D−1)∘Conv(1), wherein D represents the first number of dimensions.

### Claim 24 (depends on 17)

The method of claim 17, further comprising:
iteratively convolving a plurality of variations of the tensor using the convolution function according to an equation to convolve the variant of the tensor, wherein the equation is Conv(D)≙Conv(1)∘Conv(D−1), wherein D represents the first number of dimensions.

### Claim 25 (depends on 17)

The method of claim 17,
wherein the second number of dimensions is 2, and
wherein the first number of dimensions is at least 3.

### Claim 26 (depends on 17)

The method of claim 17,
wherein the second number of dimensions is 3, and
wherein the first number of dimensions is at least 4.

### Claim 27 (depends on 17)

The method of claim 17, further comprising:
receiving image data, wherein the tensor is associated with the image data; and
processing the image data to generate processed image data, wherein the processed tensor is associated with the processed image data.

### Claim 28 (depends on 17)

The method of claim 17, further comprising:
receiving video data, wherein the tensor is associated with the video data; and
processing the video data to generate processed video data, wherein the processed tensor is associated with the processed video data.

### Claim 29 (depends on 17)

The method of claim 17, further comprising:
receiving audio data, wherein the tensor is associated with the audio data; and
processing the audio data to generate processed audio data, wherein the processed tensor is associated with the processed audio data.

### Claim 30 (depends on 17)

The method of claim 17, further comprising:
receiving text data, wherein the tensor is associated with the text data; and
processing the text data to generate processed text data, wherein the processed tensor is associated with the processed text data.
