Algorithm 1 Learning algorithm for dynamic synapse model

Input: matrix of preprocessed input objects, initial distribution of weights, neuron parameters, plasticity parameters, vector of classes. Parameter: N_epochs, Adam optimizer, BSE loss function, threshold, patience. Output: distribution of weights of the neural network, spike times of the output. 1.Initialize a stimulus generator and define the input dimension of the feature vector. 2.Create a neural network model based on the dynamic synapse model. 3.Define the loss function and optimizer. 4.Initialize tracking variables for loss and d-prime. 5.for in $N\_epochs$ do 6.   Generate train batch. 7.   Execute a forward pass within the model. 8.   Compute the loss function by comparing the predicted values with the target. 9.   Perform backpropagation. 10.end for 11.Append the loss and d-prime values to the corresponding lists for tracking. 12.if $d-prime<threshold$ 13.    Reset the wait_count. 14.else 15.    Increase the wait_count. 16.    if $wait\_count\ge patience$ 17.           Break the training loop 18.     end if 19.end if 20.Finish the algorithm execution.