Systems and methods for tractable space optimization

Patent metadata

Field	Value
Publication number	US-2024256910-A1
Application number	US-202318104136-A
Country	US
Kind code	A1
Filing date	Jan 31, 2023
Priority date	Jan 31, 2023
Publication date	Aug 1, 2024
Grant date	—

Abstract

Systems and methods of optimizing digitally-represented space is disclosed. A request to optimize a digitally-represented space is received. The request includes a data structure storing the digitally-represented space and at least one optimization parameter. A set of elements for insertion into the digitally-represented space is obtained. Each element in the set of elements includes at least one independent variable. A predicted function for the digitally-represented space is generated that represents a relationship between the at least one optimization parameter and the at least one independent variable. The predicted function is generated by a scaled neural multiplicative model (SNMM) prediction model. An optimal allocation of a subset of the set of elements in the digitally-represented space is generated that maximizes the at least one optimization parameter. The data structure storing the digitally-represented space is updated to include the optimal allocation of the subset of the set of elements.

First claim

What is claimed is:

1 . A system, comprising:
a database; a processor communicatively coupled to the database, wherein the processor is configured to read a set of instructions to:
receive a request to optimize a digitally-represented space, wherein the request includes a data structure storing the digitally-represented space and at least one optimization parameter;
obtain, from the database, a set of elements for insertion into the digitally-represented space, wherein each element in the set of elements includes at least one independent variable;
generate a predicted function for the digitally-represented space, wherein the predicted function represents a relationship between the at least one optimization parameter and the at least one independent variable, and wherein the predicted function is generated by a scaled neural multiplicative model (SNMM) prediction model;
generate an optimal allocation of a subset of the set of elements in the digitally-represented space, wherein the optimal allocation maximizes the at least one optimization parameter; and
update the data structure storing the digitally-represented space to include the optimal allocation of the subset of the set of elements, wherein the updated data structure is stored in the database.

2 . The system of claim 1 , wherein the SNMM prediction model includes at least two linear layers each having a scaling weight and a bias.

3 . The system of claim 2 , wherein the SNMM prediction model generates the prediction function according to one or more subdivisions of the digitally-represented space.

4 . The system of claim 3 , wherein the optimal allocation is generated by an optimization process including an index of feature pairs.

5 . The system of claim 3 , wherein the digitally-represented space is representative of a retail space, and wherein the optimization parameter includes expected sales, the at least one independent parameter includes a fixture count, and wherein the SNMM model is configured to determine space optimization of a particular category within the retail space for a particular brand within a particular department.

6 . The system of claim 3 , wherein the optimal allocation is defined by the at least one independent variable, wherein the independent variable includes a set of features, and wherein the optimal allocation is constrained by an upper bound and a lower bound of the at least one independent variable.

7 . The system of claim 1 , wherein the optimal allocation of the subset of the set of elements is generated by applying a power cone formulation.

8 . The system of claim 1 , wherein the processor reads the set of instructions to generate an interface including the updated data structure storing the digitally-represented space and the optimal allocation of the subset of the set of elements.

9 . A computer-implemented method, comprising
receiving a request to optimize a digitally-represented space, wherein the request includes a data structure storing the digitally-represented space and at least one optimization parameter; obtaining, from a database, a set of elements for insertion into the digitally-represented space, wherein each element in the set of elements includes at least one independent variable; generating a predicted function for the digitally-represented space, wherein the predicted function represents a relationship between the at least one optimization parameter and the at least one independent variable, and wherein the predicted function is generated by a scaled neural multiplicative model (SNMM) prediction model; generating an optimal allocation of a subset of the set of elements in the digitally-represented space, wherein the optimal allocation maximizes the at least one optimization parameter; and updating the data structure storing the digitally-represented space to include the optimal allocation of the subset of the set of elements, wherein the updated data structure is stored in the database.

10 . The computer-implemented method of claim 9 , wherein the SNMM prediction model includes at least two linear layers each having a scaling weight and a bias.

11 . The computer-implemented method of claim 10 , wherein the SNMM prediction model generates the prediction function according to one or more subdivisions of the digitally-represented space.

12 . The computer-implemented method of claim 10 , wherein the optimal allocation is generated by an optimization process including an index of feature pairs.

13 . The computer-implemented method of claim 10 , wherein the digitally-represented space is representative of a retail space, and wherein the optimization parameter includes expected sales, the at least one independent parameter includes a fixture count, and wherein the SNMM model is configured to determine space optimization of a particular category within the retail space for a particular brand within a particular department.

14 . The computer-implemented method of claim 10 , wherein the optimal allocation is defined by the at least one independent variable, wherein the independent variable includes a set of features, and wherein the optimal allocation is constrained by an upper bound and a lower bound of the at least one independent variable.

15 . The computer-implemented method of claim 9 , wherein the optimal allocation of the subset of the set of elements is generated by applying a power cone formulation.

16 . The computer-implemented method of claim 9 , comprising generating an interface including the updated data structure storing the digitally-represented space and the optimal allocation of the subset of the set of elements.

17 . A non-transitory computer-readable medium having instructions stored thereon, wherein the instructions, when executed by at least one processor, cause a device to perform operations comprising:
receiving a request to optimize a digitally-represented space, wherein the request includes a data structure storing the digitally-represented space and at least one optimization parameter; obtaining, from a database, a set of elements for insertion into the digitally-represented space, wherein each element in the set of elements includes at least one independent variable; generating a predicted function for the digitally-represented space, wherein the predicted function represents a relationship between the at least one optimization parameter and the at least one independent variable, and wherein the predicted function is generated by a scaled neural multiplicative model (SNMM) prediction model; generating an optimal allocation of a subset of the set of elements in the digitally-represented space, wherein the optimal allocation maximizes the at least one optimization parameter; and updating the data structure storing the digitally-represented space to include the optimal allocation of the subset of the set of elements, wherein the updated data structure is stored in the database.

18 . The non-transitory computer readable medium of claim 17 , wherein the optimal allocation is defined by the at least one independent variable, wherein the independent variable includes a set of features, and wherein the optimal allocation is constrained by an upper bound and a lower bound of the at least one independent variable.

19 . The non-transitory computer readable medium of claim 18 , wherein the SNMM prediction model includes at least two linear layers each having a scaling weight and a bias, and wherein the SNMM prediction model generates the prediction function according to one or mor

Assignees

• Walmart Apollo Llc

Inventors

• Giri Drishya
• Kuhn Kenneth David
• Chiruvelli Aravind

Classifications

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N5/022 Primary

  Knowledge engineering; Knowledge acquisition · CPC title

• G06N3/04 Primary

  Architecture, e.g. interconnection topology · CPC title