# US 2021/0065004 A1 · Method and device for subnetwork sampling, and method and device for building a hypernetwork topology

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

- Number: US 2021/0065004 A1
- Document ID: US20210065004A1
- Status: Published application
- Kind: A1
- Application number: 16/689,537
- Filed: 2019-11-20
- Published: 2021-03-04
- Earliest claimed: 2019-08-29
- Assignees: Beijing Xiaomi Moble Software Co., Ltd.
- Applicants: Beijing Xiaomi Moble Software Co., Ltd.
- Inventors: Xiangxiang CHU; Ruijun XU; Bo ZHANG; Jixiang LI; Qingyuan LI; Bin WANG
- Web page: https://patentgrep.com/patent/US20210065004A1

## CPC

- G06N 3/045: Combinations of networks
- G06N 3/082: modifying the architecture, e.g. adding, deleting or silencing nodes or connections
- G06N 3/0985: Hyperparameter optimisation; Meta-learning; Learning-to-learn
- G06N 20/00: Machine learning
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A method for subnetwork sampling is applicable to a hypernetwork topoloyly. The hypernetwork topology includes n layers, each layer includes at least two substructures, and each substructure includes hatch normalization (BN) modules in one-to-one correspondence with the substructures of a closest upper layer, n>0 and n being a positive integer. The method includes: a substructure A(N) of an N-th layer is selected, 1>N≥n; a selected substructure A(N-1) of an (N−1)-th layer is determined; a BN module C(B) in one-to-to correspondence with A(N-1) is determined from the substructure A(N); and the substructure A(N) is added into a subnetwork through the BN module C(B).

## Claims

### Claim 1

A method for subnetwork sampling, applicable to a hypernetwork topology, the hypernetwork topology comprising n layers, each layer comprising at least two substructures, each substructure comprising a batch normalization (BN) module in one-to-one correspondence with a substructure of a closest upper layer, n>0 and n being a positive integer, the method comprising:
selecting a substructure A(N) of an N-th layer, 1>N≥n;
determining a selected substructure A(N-1) of an (N−1)-th layer;
determining, from the substructure A(N), a BN module C(B) in one-to-one correspondence with the substructure A(N-1); and
adding the substructure A(N) into a subnetwork through the BN module C(B).

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
selecting a substructure A(1) of a first layer.

### Claim 3 (depends on 1)

The method of claim 1, wherein adding the substructure A(N) into the subnetwork through the BN module C(B) comprises:
connecting the BN module C(B) with the substructure A(N-1) to add the substructure A(N) into the subnetwork.

### Claim 4 (depends on 2)

The method of claim 2, wherein adding the substructure A(N) into the subnetwork through the BN module C(B) comprises:
connecting the BN module C(B) with the substructure A(N-1) to add the substructure A(N) into the subnetwork.

### Claim 5 (depends on 3)

The method of claim 3, wherein the substructure A(N-1) comprises an Output module configured to output a characteristic; and connecting the BN module C(B) with the substructure A(N) further comprises:
connecting the BN module C(B) with the output module of the substructure A(N-1).

### Claim 6 (depends on 4)

The method of claim 4, wherein the substructure A(N-1) comprises an output module configured to output a characteristic; and connecting the BN module C(B) with the substructure A(N) further comprises:
connecting the BN module C(B) with the output module of the substructure A(N-1).

### Claim 7

A method for building a hypernetwork topology, comprising:
building an n-layer structure, n≥0 and n being a positive integer;
arranging in substructures in each layer, m>0;
for each layer, an N-th layer, of a second layer to an n″-th layer, arranging m batch normalization (BN) modules in each substructure; and
for each layer, the N-th layer, of the second layer to the n-th layer, establishing a one-to-one correspondence between each BN module and a substructure of an (N−1)-th layer.

### Claim 8

A device for subnetwork sampling, applicable to subnetwork sampling in a hypernetwork topology, the hypernetwork topology comprising n layers, each layer comprising at least two substructures, each substructure comprising a batch normalization (BN) module in one-to-one correspondence with a substructure of a closest upper layer, n>0 and n being a positive integer, the device comprising:
a memory configured to store an instruction; and
a processor configured to execute the instruction stored in the memory to:
select a substructure A(N) of an N-th layer, 1>N≥n;
determine a selected substructure A(N−1) of an (N−1)-th layer;
determine, from the substructure A(N), a BN module C(B) in one-to-one correspondence with the substructure A(N-1); and
add the substructure A(N) into a subnetwork through the BN module C(B).

### Claim 9 (depends on 8)

The device of claim 8, wherein the processor is further configured to select a substructure A(1) of a first layer.

### Claim 10 (depends on 8)

The device of claim 8, wherein the processor is further configured to connect the BN module C(B) with the substructure A(N-1) to add the substructure A(N) into the subnetwork.

### Claim 11 (depends on 9)

The device of claim 9, wherein the processor is further configured to connect the BN module C(B) with the substructure A(N-1) to add the substructure A(N) into the subnetwork.

### Claim 12 (depends on 10)

The device of claim 10, wherein the substructure A(N-1) comprises an output module configured to output a characteristic; and the processor is further configured to connect the BN module C(B) with the output module of the substructure A(N-1).

### Claim 13 (depends on 11)

The device of claim 11, wherein the substructure A(N-1) comprises an output module configured to output a characteristic; and the processor is further configured to connect the BN module C(B) with the output module of the substructure A(N-1).

### Claim 14

A device, comprising:
a memory configured to store an instruction; and
a processor configured to execute the instruction stored in the memory to:
build an n-layer structure, n>0 and n being a positive integer;
arrange m substructures in each layer, m>0;
for each layer, an N-th layer, of a second layer to an n-th layer, arranging in batch normalization (BN) modules in each substructure; and
for each layer, the N-th layer, of the second layer to the n-th layer, establishing a one-to-one correspondence between each BN module and a substructure of an (N−1)-th layer.

### Claim 15 (depends on 1)

A computer-readable storage medium having stored thereon an instruction that, when executed by a processor of a device, causes the device to perform the method for subnetwork sampling of claim 1.

### Claim 16 (depends on 2)

A computer-readable storage medium having stored thereon an instruction that, when executed by a processor of a device, causes the device to perform the method for subnetwork sampling of claim 2.

### Claim 17 (depends on 3)

A computer-readable storage medium having stored thereon an instruction that, when executed by a processor of a device, causes the device to perform the method for subnetwork sampling of claim 3.

### Claim 18 (depends on 5)

A computer-readable storage medium, having stored thereon an instruction that, when executed by a processor of a device, causes the device to perform the method for subnetwork sampling of claim 5.

### Claim 19 (depends on 7)

A computer-readable storage medium, having stored thereon an instruction that, when executed by a processor of a device, causes the device to perform the method for building a hypernetwork topology of claim 7.
