"Three different collate-functions are present:\n",
"- `collate_pack` packs timeseries in to a PyTorch `PackedSequence` object, which allows masking in recurrent networks\n",
"- `collate_crop` crops timeseries to the length of the shortest series in the batch, data points are removed from the front as later data points likely contribute more information toward the classification goal\n",
"- `collate_pad` pads sequences to the length of the longest sequence in the batch\n",
"\n",
"Experiments showed that `collate_crop` seems to work best for CNNs."
-`collate_pack` packs timeseries in to a PyTorch `PackedSequence` object, which allows masking in recurrent networks
-`collate_crop` crops timeseries to the length of the shortest series in the batch, data points are removed from the front as later data points likely contribute more information toward the classification goal
-`collate_pad` pads sequences to the length of the longest sequence in the batch
Experiments showed that `collate_crop` seems to work best for CNNs.