We will use the adam (Adaptive Moment Optimization) optimizer instead of the rmsprop (Root Mean Square Propagation) optimizer that we used earlier when compiling the model. To make a comparison of model performance easier, we will keep everything else the same as earlier, as shown in the following code:
# Model architecturemodel <- keras_model_sequential() %>% layer_embedding(input_dim = 500, output_dim = 32) %>% layer_lstm(units = 32) %>% layer_dense(units = 1, activation = "sigmoid")# Compilemodel %>% compile(optimizer = "adam", loss = "binary_crossentropy", metrics = c("acc"))# Fit modelmodel_two <- model %>% fit(train_x, train_y, epochs = 10, batch_size = 128, validation_split = 0.2)plot(model_two) ...
