Sparse Vector Machine Learning Model Pruning for Content Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for determining click probabilities and selecting content for presentation require significant memory and computational resources, leading to inefficiencies in storage and prediction times, and often sacrifice accuracy by omitting relevant features to meet storage requirements.
Innovation Solution
The implementation of pruning operations to generate machine learning models with sparse vector representations and weights, reducing memory usage and computational complexity while maintaining accuracy by selectively removing unnecessary feature parameters and weights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models use dense vector representations with all features, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts and removes unnecessary features and weights from the machine learning model through pruning operations. This extraction process eliminates redundant elements while preserving the essential features needed for accurate click probability determination, thereby reducing model complexity without significantly compromising measurement precision
Solution Approach 2:
The patent applies different quality standards to different parts of the model by performing selective pruning on specific features and weights based on their importance. Rather than uniformly reducing the model, it maintains high precision in critical areas while reducing complexity in less important areas, achieving local optimization of the quality-complexity tradeoff
2Measurement precision
If machine learning models include all relevant features, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent extracts and removes redundant features from the model input through pruning operations. By eliminating features that contribute minimally to prediction accuracy, the model processes fewer inputs during inference, thereby reducing prediction time while maintaining measurement precision for the remaining essential features
Solution Approach 2:
The patent applies partial action by including only the necessary subset of features rather than all possible features. The pruning process identifies and retains only the critical features needed for accurate click probability determination, excluding excess features that would increase processing time without providing proportional accuracy benefits
3Measurement precision
If machine learning models are trained with complete feature sets, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts and removes unnecessary features from the training data and model structure through pruning operations. This extraction reduces the amount of data that needs to be stored and processed during training, thereby reducing device complexity and storage requirements while preserving the essential features needed for accurate content selection
Solution Approach 2:
The patent applies selective pruning to different features based on their importance to content selection accuracy. Critical features are retained with full detail to maintain measurement precision, while less important features are reduced or removed, achieving local optimization that balances storage requirements with accuracy
Data Source
AI summary
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 sparse vector representations associated with features of the plurality of sets of 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. A plurality of positive signal probabilities associated with a plurality of content items may be determined using the machine learning model based upon one or more sparse vector representations, 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 plurality of positive signal probabilities.