Neurosynaptic Core Circuits for Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neuromorphic and synaptronic computation systems face challenges in effectively extracting features from high-dimensional data, such as images and videos, using traditional digital models, which limits their ability to mimic biological brain functionality and learn efficiently.
Innovation Solution
A neurosynaptic system comprising multiple core circuits that simulate biological neurons and axons, utilizing synaptic connectivity and spike-timing dependent plasticity to extract and combine features from input data, allowing for efficient feature extraction and scene understanding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional digital models are used for feature extraction, then the system can process data using conventional computing methods, but the ability to mimic biological brain functionality and learn efficiently is limited
Solution Approach 1:
The patent replaces traditional digital computing models with a neurosynaptic system that uses artificial neurons and synapses to process information. This substitution enables the system to mimic biological brain functionality through spiking neural networks and spike-timing dependent plasticity, fundamentally changing how computation is performed from conventional digital operations to biologically-inspired neural processing
Solution Approach 2:
The system changes the fundamental parameters of computation by using continuous-time spiking neurons instead of discrete digital values. The synaptic weights are updated based on spike timing differences rather than traditional gradient descent, and features are extracted through temporal integration of spikes rather than conventional matrix operations, enabling adaptive learning while maintaining biological fidelity
2Measurement precision
If multiple core circuits are used to extract features from different input regions, then feature extraction capability is improved, but system complexity increases
Solution Approach 1:
The system divides the input data into multiple input regions and processes each region using dedicated core circuits. Each core circuit extracts features from its assigned input region independently, allowing parallel processing of different spatial locations. This segmentation enables comprehensive feature extraction across the entire input while maintaining modular architecture
Solution Approach 2:
The patent combines features extracted from multiple input regions by integrating the outputs of different core circuits. The synthesized features are generated by combining regional features through weighted summation or other integration methods, creating a unified representation that captures both local and global patterns from the entire input data
3Productivity
If synaptic connectivity information is used to combine features, then learning efficiency is improved, but computational requirements increase
Solution Approach 1:
The system performs self-learning through spike-timing dependent plasticity, where synaptic weights are automatically adjusted based on the temporal relationship between pre-synaptic and post-synaptic spikes. This self-service mechanism eliminates the need for external training interventions or manual weight adjustments, enabling the system to improve its feature extraction capabilities autonomously through exposure to input data
Solution Approach 2:
The patent pre-establishes the framework for learning by implementing fixed synaptic connectivity patterns that guide feature combination. The synaptic weights are initialized with specific patterns that predispose the system to extract certain types of features, and learning occurs through incremental adjustments to these pre-established connections rather than building the entire system from scratch
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system enables efficient processing and learning of high-dimensional data, facilitating scene understanding, object classification, and pattern recognition by mimicking biological brain functionality, improving feature extraction and combination processes.
Implementation Method 1
The synaptic conductance changes with time as a function of the relative spike times of pre-synaptic and post-synaptic neurons, as per spike-timing dependent plasticity (STDP). Specifically, under the STDP rule, the conductance of a synapse increases if its post-synaptic neuron fires after its pre-synaptic neuron fires, and the conductance of a synapse decreases if the order of the two firings is reversed.
Data Source
AI summary
Embodiments of the invention provide a neurosynaptic system comprising a first set of one or more neurosynaptic core circuits configured to receive input data comprising multiple input regions, and extract a first set of features from the input data. The features of the first set are computed based on different input regions. The system further comprises a second set of one or more neurosynaptic core circuits configured to receive the first set of features, and generate a second set of features by combining the first set of features based on synaptic connectivity information of the second set of core circuits.