Condensing Transition Tables for Pattern Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The inefficiency of storing and using finite automata for pattern recognition due to high storage requirements, as they need to store transitions for every possible next input element for every possible state, leading to significant storage overhead.
Innovation Solution
The method involves creating condensed representations of transition tables by identifying similar transition table templates and storing differences, allowing for reduced storage overhead and improved caching performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If transition tables are stored for every possible state and input element, then pattern matching accuracy is maintained, but storage requirements increase significantly
Solution Approach 1:
The patent merges similar transition tables by identifying templates and consolidating their representations. Instead of storing separate transition tables for each state, the system merges tables that share common patterns, storing only the unique differences from the template. This reduces the total quantity of stored data while preserving the ability to accurately determine next states for pattern matching.
Solution Approach 2:
The transition table template serves multiple functions: it acts as a base representation for multiple similar transition tables, provides the common structure shared by grouped states, and enables efficient storage by representing entire groups of tables with a single template plus minimal difference data. This universal template approach maintains pattern matching accuracy while significantly reducing storage requirements.
2Quantity of substance
If condensed representations with templates are used, then storage requirements are reduced, but complexity of determining next state increases
Solution Approach 1:
The patent segments the transition table storage into two distinct parts: a template component containing the common structure and a difference component containing only the unique variations. This segmentation allows the system to store condensed representations that are simpler in structure than full transition tables, while the segmentation itself provides a clear organizational framework that manages complexity rather than increasing it.
Solution Approach 2:
The patent extracts only the essential differences from similar transition tables and stores them separately from the common template. By taking out and storing only the unique variations rather than complete tables, the system reduces storage requirements and simplifies the overall structure, as the template provides the bulk of the information and the extracted differences are minimal.
3Productivity
If similar transition tables are grouped using templates, then caching performance improves, but time to identify difference region increases
Solution Approach 1:
The patent performs preliminary actions by pre-grouping similar transition tables into templates and pre-identifying difference regions during the table creation or loading phase. This preliminary organization allows the system to store pre-computed difference regions that can be quickly retrieved during pattern matching, rather than computing differences in real-time. The preliminary grouping and difference identification enables faster caching performance without significant overhead during actual pattern matching operations.
Data Source
Figure 1A
Figure 1B
Figure 2A
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for using transition table templates. In one aspect, a method includes receiving a transition table for a current state of a finite automaton and determining whether the transition table for the current state is similar to a transition table template in a set of transition table templates. The method further includes generating a condensed representation of the transition table if the transition table is similar to a transition table template and otherwise adding the transition table to the set of transition table templates. In another aspect, a method includes receiving an input element and determining whether a next state corresponding to the input element is in the difference region of a condensed transition table. The method further includes retrieving the next state from the difference region, or a transition table template, based on the determination.