Associative Memory Rank-Based Learning for Pattern Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pattern recognition systems using associative memory face challenges in efficiently learning and recognizing new reference data, leading to prolonged processing times and lack of effective learning methods for replacing or adding reference data.
Innovation Solution
A reference data recognition learning method that employs a rank-based system within an associative memory to prioritize and update reference data, comparing input search data with stored data to determine similarity and adjust ranks accordingly, allowing for automatic learning of new data and efficient pattern recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If sequential comparison method is used to compare input pattern data with reference data, then the processing can be implemented with simple memory structure, but much processing time is taken and no learning concept exists
Solution Approach 1:
The memory is segmented into multiple banks (first memory bank and second memory bank) to enable parallel processing. The first memory bank stores first reference data while the second memory bank stores second reference data, allowing simultaneous comparison operations that reduce processing time while maintaining structured memory organization.
Solution Approach 2:
The system dynamically switches between different reference data sets stored in different memory banks based on recognition needs. The control unit dynamically selects which memory bank to access, enabling adaptive processing that reduces time loss while maintaining simple memory structure through dynamic rather than static reconfiguration.
2Device complexity
If reference data is sequentially compared, then the memory structure remains simple, but the processing time increases in proportion to the increase of the number of comparison data
Solution Approach 1:
The reference data is segmented and distributed across multiple memory banks, allowing parallel comparison operations. This segmentation enables the system to process multiple reference data items simultaneously, increasing recognition speed without requiring complex memory restructuring.
Solution Approach 2:
The system uses partial parallel processing by comparing reference data from different memory banks simultaneously. This partial action approach increases productivity by processing multiple items in parallel while keeping the overall memory structure relatively simple, avoiding the need for complete system redesign.
3Adaptability or versatility
If neural network is used to realize pattern recognition and learning function, then learning capability is achieved, but there is no preferable method of effectively storing patterns and relearning is required throughout the whole network
Solution Approach 1:
The learning function is segmented and implemented in the control unit rather than throughout the entire network. The control unit manages rank assignments and data storage decisions, separating the learning control function from the pattern storage function. This reduces network structure complexity while maintaining learning capability through centralized coordination.
Solution Approach 2:
The control unit acts as an intermediary between the pattern recognition function and the memory storage function. It mediates by assigning ranks to reference data and determining which data to store in which memory banks, enabling learning without requiring relearning throughout the entire network structure.
4Speed
If associative memory with fully parallel architecture is used, then high-speed pattern matching is achieved, but an effective learning method of recognizing new reference data is not established
Solution Approach 1:
The system dynamically assigns ranks to reference data based on their similarity to input patterns. The control unit dynamically determines which reference data should be stored in which memory banks, enabling the high-speed associative memory to adapt to new learning requirements without sacrificing its parallel processing speed.
Solution Approach 2:
The system uses feedback from pattern recognition results to adjust the ranking and storage of reference data. When patterns are recognized, the control unit uses this information to update the rank assignments, creating a feedback loop that enables continuous learning while maintaining high-speed operation through the established parallel architecture.
Data Source
AI summary
In an associative memory, when a reference data having the minimum distance with respect to an input data is detected as winner, it is determined whether or not a distance between the input data and winner is less than a threshold value. If the distance is less than the threshold value, it is determined that the reference data detected as winner matches with the input data, and then, a rank of the reference data is improved. If the distance is more than the threshold value, it is determined that the reference data is data different from the input data, and then, the input data is written as new reference data to the associative memory and replaces the reference data with the lowest rank. The upper positions of rank form as a long-term memory, and the lower positions thereof form as a short-term memory.