“…We sequentially compress the input data into various bottleneck dimensions (k) from 2 dimensions to 200 dimensions. We use k = 2, 3, 4, 5,6,7,8,9,10,12,14,16,18,20,25,30,35,40,45,50,60,70,80,90,100,125,150, and 200 for a total of 28 different dimensions. For each model, we train five independent times using five different random seed initializations.…”