Filtering Co-occurrence Data via Neural Network Intermediary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models face challenges in accurately predicting future events due to noise in co-occurrence data, which affects the reliability of event predictions and recommendations.
Innovation Solution
The approach involves training an artificial neural network (ANN) using encoded representations of past events to determine similarities and differences, filtering noise by using a distance function like cosine distance to modify or remove unlikely co-occurrence data, and training a second neural network to generate likelihoods of future events based on these analyses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If co-occurrence data is used for event prediction, then prediction coverage is improved, but noise in the data reduces prediction accuracy
Solution Approach 1:
The patent introduces an intermediate representation layer that transforms raw co-occurrence data into a filtered, structured format before feeding it to the prediction model. This intermediary processing step separates the useful signal from noise while preserving the comprehensive event relationships, thereby maintaining prediction coverage while improving accuracy.
Solution Approach 2:
The patent extracts and removes noisy or irrelevant co-occurrence patterns from the data while retaining meaningful relationships. By selectively extracting high-quality signals from the raw co-occurrence data, the system maintains broad prediction coverage while eliminating accuracy-reducing noise.
2Speed
If traditional neural network forward pass is used, then computation speed is maintained, but prediction accuracy is limited by noisy data
Solution Approach 1:
The patent performs data filtering, noise removal, and intermediate representation creation before the neural network forward pass. By preparing clean, structured input data in advance, the system maintains fast computation speeds during inference while significantly improving prediction accuracy through pre-processed high-quality inputs.
Data Source
AI summary
The present disclosure is directed to filtering co-occurrence data. In one embodiment, a machine learning model can be trained. An output of an intermediate structure of the machine learning model (e.g., an output of an internal layer of a neural network) can be used as a representation of an event. Similarities between representations of events can be determined and used to generate, augment, or modify co-occurrence data.