Spiking Neural Network Device Learning Blank Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spiking neural networks struggle to learn information with low spike density, such as blank data, due to unchanged synaptic weights, which limits their ability to recognize patterns in images with empty pixels, and existing solutions require doubling the hardware and energy consumption.
Innovation Solution
A spiking neural network device with a synaptic element, neuron circuit, synaptic potentiator, synaptic depressor, and determinator that updates synaptic weights based on firing frequency instead of sum, allowing for synaptic normalization without increasing hardware size or energy requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If synaptic normalization is implemented to enable learning of blank data, then learning capability for low spike density information is improved, but hardware size and energy consumption increase
Solution Approach 1:
The patent changes the parameter used for synaptic weight update from spike density to firing frequency. This parameter change allows the system to learn blank data without requiring synaptic normalization hardware, thus improving learning capability while avoiding increased hardware complexity
Solution Approach 2:
The patent uses a simplified copying mechanism where the firing frequency directly updates synaptic weights through potentialiation and depression operations, avoiding the need to copy and process complex spike density information across the network
2Adaptability or versatility
If synaptic normalization is implemented to enable learning of blank data, then learning capability for low spike density information is improved, but energy consumption increases
Solution Approach 1:
By changing the update parameter from spike density to firing frequency, the system achieves blank data learning through simpler frequency-based potentialiation and depression operations, significantly reducing the computational energy required compared to spike density-based normalization
Solution Approach 2:
The patent extracts only the essential firing frequency information needed for learning blank data, discarding the computationally expensive spike density calculation and normalization steps, thereby reducing energy consumption while maintaining learning capability
3Measurement precision
If STDP learning is used to update synaptic weights, then learning of spike timing information is improved, but learning of blank data fails due to unchanged synaptic weights
Solution Approach 1:
The patent introduces dynamic potentialiation and depression operations that activate based on firing frequency thresholds. This dynamic mechanism allows synaptic weights to change not only in response to spike timing but also in response to firing patterns, enabling learning of both precise timing information and blank data
Solution Approach 2:
The determinator performs preliminary assessment of firing frequency before triggering weight updates. This preliminary action ensures that synaptic weights are updated appropriately for blank data patterns before STDP learning occurs, preventing the failure to learn low spike density information
Data Source
AI summary
A spiking neural network device according to an embodiment includes a synaptic element, a neuron circuit, a determinator, a synaptic depressor, and a synaptic potentiator. The synaptic element has a variable weight and outputs, in response to input of a first spike signal, a synaptic signal having intensity adjusted in accordance with the weight. The neuron circuit outputs a second spike signal in a case where the synaptic signal is inputted and a predetermined firing condition for the synaptic signal is satisfied. The determinator determines whether or not the weight is to be updated on a basis of an output frequency of the second spike signal by the neuron circuit. The synaptic depressor performs depression operation for depressing the weight in a case where it is determined that the weight is to be updated. The synaptic potentiator performs potentiating operation for potentiating the weight.