Spiking Neural Network Autonomous Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pattern recognition systems, particularly those using deep convolutional neural networks, are inflexible when dealing with unknown patterns or rapidly changing feature templates, and struggle with real-time machine learning and autonomous feature extraction.
Innovation Solution
A system comprising a first spiking neural network that autonomously learns to recognize patterns through spike timing dependent plasticity (STDP) and lateral inhibition, and a second neural network that labels these patterns, enabling rapid real-time machine learning and feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep convolutional neural networks are used for pattern recognition, then accuracy in recognizing known patterns is improved, but flexibility when dealing with unknown patterns or rapidly changing feature templates deteriorates
Solution Approach 1:
The patent implements dynamic adaptability by enabling the neural network to continuously learn and update its feature templates in real-time through streaming data, allowing the system to adapt to unknown patterns and rapidly changing feature templates while maintaining recognition accuracy
Solution Approach 2:
The system performs autonomous feature extraction and template updates without requiring manual retraining or intervention. The neural network automatically learns new patterns and adapts to changing conditions through continuous processing of sensory data streams, making the system self-adjusting and flexible
2Ease of operation
If traditional neural networks with manual training are used, then control over feature extraction is improved, but real-time machine learning capability deteriorates
Solution Approach 1:
The neural network autonomously performs feature extraction and template learning in real-time without manual intervention. The system automatically processes sensory data streams, extracts relevant features, and updates templates dynamically, achieving both operational simplicity and real-time learning capability
Solution Approach 2:
The system continuously learns and processes patterns in real-time through continuous data streams rather than batch processing. This continuous learning operation enables real-time machine learning while maintaining control over the feature extraction process through the network's architecture and learning rules
3Reliability
If deep convolutional networks are used for feature extraction, then recognition performance on known patterns is improved, but adaptability to changing environments deteriorates
Solution Approach 1:
The system dynamically adapts to changing environments by continuously updating feature templates through real-time processing of sensory data streams. The neural network adjusts its internal representations and learning parameters on-the-fly, maintaining reliable recognition performance across varying conditions without requiring manual reconfiguration
Data Source
AI summary
Embodiments of the present invention provide an artificial neural network system for feature pattern extraction and output labeling. The system comprises a first spiking neural network and a second spiking neural network. The first spiking neural network is configured to autonomously learn complex, temporally overlapping features arising in an input pattern stream. Competitive learning is implemented as spike timing dependent plasticity with lateral inhibition in the first spiking neural network. The second spiking neural network is connected with the first spiking neural network through dynamic synapses, and is trained to interpret and label the output data of the first spiking neural network. Additionally, the labeled output of the second spiking neural network is transmitted to a computing device, such as a central processing unit for post processing.