Correction Circuit for Function Approximation Outliers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data processing systems face challenges in efficiently correcting function approximation outliers, particularly in achieving accurate results while minimizing computational expense, especially in applications like AI where accuracy is less critical but tolerance is essential.
Innovation Solution
The development of methods and apparatus for generating a correction circuit that identifies and corrects outliers by using conjunctive/disjunctive normal form analysis and logical predicates, allowing for efficient correction of function approximation outliers with reduced computational overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If standard functions are used for AI applications, then computational speed is improved, but accuracy is degraded
Solution Approach 1:
The patent introduces an intermediary correction circuit between the standard function approximation and the final output. This correction circuit identifies and corrects outlier cases where the approximation fails, allowing the system to use fast approximations for most cases while ensuring accuracy for critical cases. The correction circuit acts as a mediator that bridges the gap between speed and accuracy requirements.
Solution Approach 2:
The patent applies local quality by providing different levels of processing precision for different input cases. Instead of uniformly high precision for all inputs, the system uses fast approximation for typical cases and applies correction only where needed. The correction circuit selectively processes only those cases that fall outside the acceptable error bounds, optimizing the trade-off between speed and accuracy locally rather than globally.
2Measurement precision
If additional steps are taken to achieve highest level of accuracy, then accuracy is improved, but computational expense increases
Solution Approach 1:
The patent implements partial action by applying correction only to the extent necessary - specifically, only to cases that fall outside acceptable error bounds. The correction circuit uses a threshold-based approach where full correction is applied only when needed, rather than always applying the most accurate but computationally expensive method. This partial application of correction reduces overall computational expense while maintaining required accuracy levels.
Solution Approach 2:
The patent extracts and separates the correction function from the main computation path. By identifying and isolating only the outlier cases that require correction, the system avoids applying expensive correction logic to all inputs. The correction circuit extracts only the necessary corrective actions for specific problematic cases, leaving the majority of computations to use the faster approximation method.
3Device complexity
If arithmetic engines share components, then device complexity is reduced, but both speed and accuracy requirements cannot be simultaneously met
Solution Approach 1:
The patent introduces dynamic adaptability into the shared arithmetic engine through the correction circuit. The system can dynamically adjust its behavior based on the input characteristics and required output quality. The correction circuit monitors the approximation results and dynamically applies correction only when necessary, allowing the shared components to serve both high-speed approximation and high-accuracy computation needs flexibly.
Solution Approach 2:
The patent enhances the universality of the shared arithmetic engine by adding a multi-functional correction capability. The correction circuit serves multiple purposes: it corrects accuracy errors, identifies outlier cases, and enables the system to meet both speed and accuracy requirements using the same hardware infrastructure. This multi-functionality allows a single shared engine to adapt to different performance requirements without requiring separate dedicated hardware for each function.
Data Source
AI summary
Correction of outliers in a data set includes receiving a first set of inputs of an input dataset requiring positive correction; and receiving a second set of inputs of the input dataset requiring negative correction. Conjunctive clauses with a predetermined number of terms that make all members in the second set of inputs false are identified to form a set of identified conjunctive clauses. Members from the first set of inputs that evaluate to true are collected for each conjunctive clause in the set of identified clauses. The set of identified conjunctive clauses are iterated through until all of the first set of inputs evaluates to true, and the conjunctive clauses are disjuncted to form a disjuncted expression. A correction circuit for the input dataset is generated based on the disjuncted expression.