Ridge Regression Model Dynamic Update for Streaming Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data models based on ridge regression methodologies are computationally inefficient when applied to dynamic, streaming data, as they require rebuilding the entire model with every new data arrival, which is time-consuming and resource-intensive.
Innovation Solution
A computing apparatus that dynamically updates the ridge regression data model by caching new data values and recalculating parameters periodically, using a Bayesian perspective to minimize time and computational resources, allowing for real-time updates of the model as new data is continuously received.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire ridge regression model is rebuilt with every new data arrival, then the model remains accurate and up-to-date, but the computational time and resource consumption increase significantly
Solution Approach 1:
The patent segments the model updating process into two distinct parts: (1) caching new data values as they arrive without processing, and (2) periodically recalculating parameters using accumulated cached data. This segmentation allows the system to maintain data freshness while avoiding continuous computationally expensive model rebuilding, thus resolving the contradiction between model accuracy and computational time.
Solution Approach 2:
The patent implements periodic parameter recalculation instead of continuous model rebuilding. New data is cached continuously, but the actual parameter recalculation occurs at periodic intervals when sufficient data has accumulated. This periodic action maintains model accuracy while dramatically reducing computational time and resource consumption compared to rebuilding the model with every new data arrival.
2Adaptability or versatility
If the entire ridge regression model is rebuilt with every new data arrival, then the model adapts to new data, but the computational resources required increase significantly
Solution Approach 1:
The patent segments the data processing workflow into data collection (caching) and data processing (parameter recalculation) phases. By separating these functions, the system can continuously adapt to new data by caching it, while resource-intensive computations are performed only periodically when resources are available, thus maintaining adaptability while reducing computational resource usage.
Solution Approach 2:
The patent performs preliminary action by caching new data values as they arrive without immediate processing. This preliminary data collection allows the system to prepare for future model updates by having data ready, but delays the computationally expensive parameter recalculation until periodic intervals, thus maintaining model adaptability while conserving computational resources.
3Speed
If data is processed continuously in real-time, then the model updates are timely, but the computational complexity increases
Solution Approach 1:
The patent implements periodic parameter recalculation as a simplified alternative to continuous real-time processing. By accumulating data in cache and performing parameter updates at periodic intervals rather than continuously, the system achieves timely model updates without the high computational complexity associated with processing every new data point immediately, thus balancing update speed with computational complexity.
Data Source
AI summary
Dynamically updating a ridge regression data model of a continuous stream of data is disclosed. New data chunks corresponding to a current data accumulation point are received and the data values in the new data chunks are transformed via standardization methods. A ridge estimator for the standardized data that includes data chunks received up to a penultimate data accumulation point to include the new data chunks is dynamically updated. The cumulative observations received up to the current data accumulation point are updated and stored. Predictions for the continuous data stream are generated based on the updated ridge estimator.