Modified Coordinate Descent for LASSO Model Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating statistical models from large data sets is computationally intensive, requiring significant processing power, memory, and time, which can slow down electronic devices and reduce available resources.
Innovation Solution
A modified coordinate descent method is used in conjunction with the least absolute shrinkage and selection operator (LASSO) method to determine regression coefficient values, reducing the number of processing iterations and memory accesses by using a subset of independent variables and employing a ray-continuation property to minimize the LASSO cost function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional statistical models are generated from large data sets using conventional methods, then model accuracy and representational capability are improved, but processing time, computational power consumption, and memory requirements increase significantly
Solution Approach 1:
The patent applies coordinate descent method which segments the optimization problem into iterative steps, each optimizing one coefficient at a time. This segmentation allows the complex model generation process to be broken down into manageable iterations, reducing overall computational burden while maintaining accuracy through progressive refinement of coefficients
Solution Approach 2:
The patent modifies the LASSO cost function by introducing a refinement factor parameter that changes the optimization landscape. This parameter transformation enables faster convergence to minimum values by adjusting the curvature and steepness of the cost function, thereby reducing processing iterations and time while maintaining model accuracy
2Measurement precision
If traditional statistical models are generated from large data sets, then model accuracy is improved, but processing power and electrical power consumption increase significantly
Solution Approach 1:
The patent uses LASSO regularization which applies partial action by selectively penalizing only the least important coefficients, effectively performing feature selection. This partial optimization approach reduces the number of meaningful parameters to process, thereby reducing processing power and electrical power consumption while maintaining sufficient model accuracy through sparsity-induced feature selection
Solution Approach 2:
By transforming the optimization problem through parameter changes in the cost function (introducing refinement factor and modified LASSO penalty terms), the patent creates a more efficient energy utilization pattern. The modified cost function converges faster with fewer iterations, directly reducing the cumulative energy consumption while preserving model accuracy
3Productivity
If traditional coordinate descent method is used to minimize LASSO cost function, then model generation is achieved, but the number of processing iterations and memory accesses is excessive
Solution Approach 1:
The patent implements feedback mechanisms through the refinement factor that adapts the optimization trajectory based on previous iterations' performance. This feedback loop allows the algorithm to learn from each iteration and adjust subsequent steps, reducing the total number of iterations needed to converge while maintaining accurate model generation
Solution Approach 2:
The patent performs preliminary actions by pre-computing and storing intermediate values and gradient information before the main optimization loop. This preliminary preparation reduces the computational workload during each iteration and decreases the total number of processing steps required, thereby improving model generation speed without sacrificing accuracy
Data Source
AI summary
Processing speeds for generating a model can be enhanced. For example, the model can be generated by using regression coefficient values as weights for independent variables in the model. The regression coefficient values can be determined using a coordinate descent method to find a minimum value of a least absolute shrinkage and selection operator cost function. Each iteration of the coordinate descent method can include determining a starting coordinate based on (i) a previous starting coordinate or a previous regression coefficient value from an immediately prior iteration of the coordinate descent method; (ii) a current regression coefficient value associated with a current iteration of the coordinate descent method; and (iii) a refinement factor configured to minimize a result of a univariate algorithm. Each iteration can also include performing a coordinate descent using the starting coordinate to determine a next regression coefficient value for a next iteration of the coordinate descent method.