Neural Network Device with Engineered Delays for Pattern Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pattern matching technologies, such as the Boyer-Moore algorithm and deterministic finite automata, are inefficient for large alphabets and require significant computational resources, especially when dealing with wildcards and partial patterns in large data sets like internet traffic and video streams.
Innovation Solution
A neural network device with engineered delays is used for pattern storage and matching, where each character in a stored pattern has a specific connection delay, allowing for efficient detection of patterns in input data streams by simulating the temporal occurrence of characters, thus enabling fast recognition and reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If traditional pattern matching algorithms (Boyer-Moore, DFA) are used, then pattern matching can be performed, but recognition speed is insufficient for large data sets and large alphabets
Solution Approach 1:
The patent transitions from traditional sequential pattern matching algorithms to a neural network architecture that processes patterns in parallel across multiple dimensions. The neural network uses distributed representations and simultaneous activation of multiple neurons to achieve faster recognition speeds, particularly for large alphabets and data sets.
Solution Approach 2:
The pattern matching task is divided into multiple independent neural units that process different aspects of the pattern simultaneously. Each neural neuron handles specific feature detection, and the collective activation of these segmented units enables rapid pattern recognition without the sequential bottlenecks of traditional algorithms.
2Measurement precision
If exact pattern matching algorithms are used, then computational precision is maintained, but computational complexity increases significantly
Solution Approach 1:
The patent changes the fundamental parameters of pattern matching by using neural network activation thresholds and weighted sums instead of traditional algorithmic comparisons. This parameter transformation allows the system to maintain matching accuracy while reducing computational complexity through parallel processing and approximate matching capabilities.
Solution Approach 2:
The patent replaces traditional mechanical/computational algorithms with a neural network system that uses biological-inspired processing. The neural network substitutes sequential computational steps with parallel neural activations, reducing the mechanical complexity of pattern matching while maintaining or improving accuracy.
3Adaptability or versatility
If state-transition tables are built for DFA, then pattern matching with wildcards is enabled, but additional preparation cost and time are required
Solution Approach 1:
The neural network performs preliminary learning and adaptation during its initialization and training phases, automatically building its internal state-transition structures. This preliminary action eliminates the need for explicit state-transition table construction at runtime, as the network has already learned the appropriate transitions during its preparatory learning phase.
Solution Approach 2:
The neural network automatically constructs and maintains its own state-transition representations through self-organization and learning mechanisms. Instead of requiring external preparation of state-transition tables, the system serves itself by dynamically adapting its internal structure to handle wildcards and variable-length patterns without additional preparation overhead.
Data Source
AI summary
Described is a system for searching a continuous data stream for exact matches with a priori stored data sequences. The system includes a neural network with an input and an output layer. The input layer has one neuron for each possible character or number in the data stream, and the output layer has one neuron for each stored pattern. Importantly, the delays of the connections from input to output layer are engineered to match the temporal occurrence of an input character within a stored sequence. Thus, if an input sequence has the proper time gaps between characters, matching a stored pattern, then the delayed neural signals result in a simultaneous activation at the receiving neuron, which indicates a detected pattern. For storing a pattern, only one connection for each pair of input character and output neuron has to be specified resulting in sparse coding and quick storage.