Fig. 10
From: Comparative analysis of deep learning architectures in solar power prediction
Residual error distributions (in kW) on the test dataset for different deep learning models, visualized with histograms and kernel density estimates: (a) Transformer, (b) LSTM, (c) GRU, (d) CNN, (e) InformerLite, (f) SimpleRNN, (g) Temporal Convolutional Network (TCN), and (h) Autoencoder.
