10K node 16 nearest neighbor bi-partitioned network with 10 pairs of complementary (opposite polarity, opposite partition) bias nodes changing polarity with different periods at T(node)/T(edge)=120. The simulation periodically applies and removes the bias potentials to variously evolve under conditions with and without input. Bias nodes are green when on and red when off. Starting from an initially disordered state, the network evolves to efficiently conduct charge between nodes and rapidly adapt to changing node polarities. When the bias is removed the network dynamics continue but are clearly influenced by the modular structure that emerged during the bias phases, resembling both the unbiased and biased stages of its evolution. Edge weight magnitudes are largely preserved during the unbiased phases of evolution even as they are continuously updated. This video is associated with Figure 12 in the Thermodynamic Neural Network manuscript.
10k Node 16NN Bi-partitioned Network with 20 Complementary Bias Nodes with On-Off Bias Phases
10K node 16 nearest neighbor bi-partitioned network with 10 pairs of complementary (opposite polarity, opposite partition) bias nodes changing polarity with different periods at T(node)/T(edge)=120. The simulation periodically applies and removes the bias potentials to variously evolve under conditions with and without input. Bias nodes are green when on and red when off. Starting from an initially disordered state, the network evolves to efficiently conduct charge between nodes and rapidly adapt to changing node polarities. When the bias is removed the network dynamics continue but are clearly influenced by the modular structure that emerged during the bias phases, resembling both the unbiased and biased stages of its evolution. Edge weight magnitudes are largely preserved during the unbiased phases of evolution even as they are continuously updated. This video is associated with Figure 12 in the Thermodynamic Neural Network manuscript.
0 Comments