Pruning field weights for content selection

Patent metadata

Field	Value
Publication number	US-12026754-B2
Application number	US-202318213914-A
Country	US
Kind code	B2
Filing date	Jun 26, 2023
Priority date	Sep 22, 2020
Publication date	Jul 2, 2024
Grant date	Jul 2, 2024

Abstract

One or more computing devices, systems, and/or methods are provided. A machine learning model may be trained using a plurality of sets of information. One or more pruning operations may be performed in association with the training to generate a machine learning model with a sparse set of field weights associated with feature fields associated with features of the plurality of sets of auction information. A request for content associated with a client device may be received. A set of features associated with the request for content may be determined. Positive signal probabilities associated with a plurality of content items may be determined using the machine learning model based upon field weights, of the machine learning model, associated with the set of features. A content item may be selected from the plurality of content items for presentation via the client device based upon the positive signal probabilities.

First claim

What is claimed is:

1. A method, comprising:
performing one or more pruning operations to generate a first machine learning model with a sparse set of field weights associated with feature fields associated with features of a plurality of sets of auction information;
receiving a bid request, wherein:
the bid request is associated with a request for content associated with a client device; and
the bid request is indicative of a set of features comprising a first feature associated with a first feature field and a second feature associated with a second feature field;

determining, using the first machine learning model, a plurality of click probabilities associated with a plurality of content items based upon one or more first field weights, of the first machine learning model, associated with the set of features;
selecting, from the plurality of content items, a content item for presentation via the client device based upon the plurality of click probabilities; and
submitting a bid value associated with the content item to an auction module for participation in an auction associated with the request for content.

2. The method of claim 1 , wherein:
the one or more pruning operations are performed in an iterative pruning process.

3. The method of claim 2 , comprising:
training a machine learning model comprising performing one or more first training steps to generate a first plurality of field weights, wherein:
the performing the one or more pruning operations comprises performing a first iteration of the iterative pruning process by setting a first subset of field weights, of the first plurality of field weights, to zero to generate a second plurality of field weights having a first sparsity;
the training the machine learning model comprises performing one or more second training steps, using the second plurality of field weights, to generate a third plurality of field weights; and
the performing the one or more pruning operations comprises performing a second iteration of the iterative pruning process by setting a second subset of field weights, of the third plurality of field weights, to zero to generate a fourth plurality of field weights having a second sparsity.

4. The method of claim 3 , wherein:
iterations of the iterative pruning process, comprising the first iteration and the second iteration, are performed until a fifth plurality of field weights is generated having a third sparsity that meets a target sparsity; and
the training the machine learning model comprises performing one or more third training steps, using the fifth plurality of field weights, to generate the sparse set of field weights.

5. The method of claim 3 , wherein:
iterations of the iterative pruning process, comprising the first iteration and the second iteration, are performed until the sparse set of field weights is generated having a third sparsity that meets a target sparsity.

6. The method of claim 3 , wherein:
the setting the first subset of field weights to zero is performed based upon a determination that field weights of the first subset of field weights are lowest field weights of the first plurality of field weights; and
the setting the second subset of field weights to zero is performed based upon a determination that field weights of the second subset of field weights are lowest field weights of the third plurality of field weights.

7. The method of claim 1 , wherein:
the one or more pruning operations are performed after training a machine learning model associated with the one or more pruning operations.

8. The method of claim 7 , wherein:
the training the machine learning model comprises generating a second machine learning model with a first plurality of field weights; and
the one or more pruning operations are performed by setting a first subset of field weights, of the first plurality of field weights, to zero to generate the sparse set of field weights.

9. The method of claim 1 , comprising:
determining the bid value based upon a first click probability of the plurality of click probabilities.

10. The method of claim 1 , wherein:
a first field weight is pruned via the one or more pruning operations; and
prior to the performing the one or more pruning operations, the first field weight is equal to a first value.

11. The method of claim 10 , wherein:
the determining the plurality of click probabilities is not performed based upon the first value.

12. A computing device comprising:
a processor; and
memory comprising processor-executable instructions that when executed by the processor cause performance of operations, the operations comprising:
performing one or more pruning operations to generate a first machine learning model with a sparse set of field weights associated with feature fields associated with features of a plurality of sets of information;
receiving a request for content associated with a client device;
determining, based upon the request for content, a set of features associated with the request for content, wherein:
the set of features comprises a first feature associated with a first feature field and a second feature associated with a second feature field;

determining, using the first machine learning model, a plurality of positive signal probabilities associated with a plurality of content items based upon one or more first field weights, of the first machine learning model, associated with the set of features;
selecting, from the plurality of content items, a content item for presentation via the client device based upon the plurality of positive signal probabilities; and
transmitting the content item to the client device.

13. The computing device of claim 12 , wherein:
the one or more pruning operations are performed in an iterative pruning process.

14. The computing device of claim 13 , the operations comprising:
training a machine learning model comprising performing one or more first training steps to generate a first plurality of field weights, wherein:
the performing the one or more pruning operations comprises performing a first iteration of the iterative pruning process by setting a first subset of field weights, of the first plurality of field weights, to zero to generate a second plurality of field weights having a first sparsity;
the training the machine learning model comprises performing one or more second training steps, using the second plurality of field weights, to generate a third plurality of field weights; and
the performing the one or more pruning operations comprises performing a second iteration of the iterative pruning process by setting a second subset of field weights, of the third plurality of field weights, to zero to generate a fourth plurality of field weights having a second sparsity.

15. The computing device of claim 14 , wherein:
iterations of the iterative pruning process, comprising the first iteration and the second iteration, are performed until a fifth plurality of field weights is generated having a third sparsity that meets a target sparsity; and
the training the machine learning model comprises performing one or more third training steps, using the fifth plurality of field weights, to generate the sparse set of field weights.

16. The computing device of claim 14 , wherein:
iterations of the iterative pruning process, comprising the first iteration and the second iteration, are performed until the sparse set of field weights is generated having a third sparsity that meets a target sparsity.

17. The computing device of claim 14 , wherein:
the setting th

Assignees

• Yahoo Assets Llc

Inventors

• Pan Junwei
• Zhou Tian
• Flores Aaron Eliasib

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/082

  modifying the architecture, e.g. adding, deleting or silencing nodes or connections · CPC title

• G06F16/90344

  by using string matching techniques · CPC title

• G06Q30/0275 Primary

  Auctions · CPC title