Reverse Engineering Neural Connectivity: Mapping Neural Activity Data to Artificial Neural Networks for Synaptic Strength Analysis

Reverse Engineering Neural Connectivity: Mapping Neural Activity Data to Artificial Neural Networks for Synaptic Strength Analysis

연구 분야: Artificial Intelligence

학회: 2024 8th International Conference on Information Technology (InCIT)

We introduce a novel approach for reverse engineering the synaptic connectivity by mapping neural activity data from live, alert zebrafish larvae to continuous-valued artificial neural networks (ANNs). By utilizing high-performance calcium indicators and advanced imaging techniques, we record real-time neuronal signals at a cellular resolution. These signals are subsequently mapped to an all-to-all connected Hopfield network with rectified linear units (ReLU) as activation functions. This ANN model can then be trained via backpropagation and other optimization techniques to obtain the synaptic strengths between the individual neurons. This approach offers considerable potential for understanding complex neural networks and brain function, and contributing to the development of more advanced brain-inspired artificial intelligence.