Fig. 5: In-plane synapse device for supervised learning. | Nature Communications

Fig. 5: In-plane synapse device for supervised learning.

From: 2D ferroelectric narrow-bandgap semiconductor Wurtzite’ type α-In2Se3 and its silicon-compatible growth

Fig. 5: In-plane synapse device for supervised learning.

a 25 cycles of repetitive LTP and LTD operations using an incremental pulse scheme. b The averaged conductance changes versus pulse number. Error bars indicate the standard deviation from which the cycle-to-cycle variations can be determined. Solid lines are the exponential fittings by using the nonlinear factor v. c The non-symmetry factors of 5 randomly selected devices. For each device, 5 cycles of LTP and LTD are averaged, and the error bars represent the standard deviation. d The designed neural network for implementing the backpropagation algorithm. e The training accuracies of WZ’ type α-In2Se3-based ANNs for the MNIST images. Ideal device-based ANN is for comparison.

Back to article page