A Lightweight Deep Learning-Based Approach for Jazz Music Generation in MIDI Format

Table 1

Results analysis of the model for different datasets.

Data set	Midi files dataset 1	Midi files dataset 2	Classical dataset
Dataset length	20 minutes	50 minutes	20 minutes
Number of epochs run	100	100	50
Time for each epoch	∼26 seconds	∼62 seconds	∼35 seconds
Batch size	16	16	16
Initial loss value	4.88	4.88	4.69
Final loss value	0.031	0.062	0.117