Skip to content

Latest commit

 

History

History
18 lines (16 loc) · 2.1 KB

todo.md

File metadata and controls

18 lines (16 loc) · 2.1 KB

  • after extractio next thing would be how to extract the files

  • we need to delete all .tgz files after extraction to save space

  • next is we need to load these .wav files individually

  • next is we need to preprocess these loaded wav files using some kind of library individually

  • take not that we do not the individual .wav files, we need to look at each .wav files as single data points

  • then we do feature extraction on these individual data points or rather single signals, based on some kind of window

  • labels are placed inside readme.md file in 5th line starting from line 1

  • calculate mel spectogram from each segment as a result of each window and calculate the necessary features from there

  • calculate zero cross rate features

  • but all these have and should be calculated strictly right after splitting the data first because this is to prevent data leakage

  • use this sample command for running training of deep learning models python tuning_dl.py -m lstm -c deep -lr 1e-3 --batch_size 32 --mode training --hyper_param_list window_time_1 hop_time_0.5 n_a_64 dense_drop_prob_0.2 rnn_drop_prob_0.2 n_units_2

  • python tuning_dl.py -m lstm -c deep -lr 1e-3 --batch_size 32 --mode training --hyper_param_list window_time_0.125 hop_time_0.0625 n_a_128 dense_drop_prob_0.2 rnn_drop_prob_0.2 n_units_2

  • python tuning_dl.py -m lstm -c deep -lr 1e-3 --batch_size 32 --mode training --hyper_param_list hertz_8000 window_time_0.25 hop_time_0.125 n_a_128 dense_drop_prob_0.2 rnn_drop_prob_0.2 - this is if number of subjects is 10

  • python tuning_dl.py -m lstm -c deep -lr 1e-3 --batch_size 256 --mode training --hyper_param_list hertz_8000 window_time_0.25 hop_time_0.125 n_a_128 dense_drop_prob_0.2 rnn_drop_prob_0.2 - this is if number of subjects is 100

  • commands for softmax regression model: python tuning_dl.py -m softmax -c trad -lr 1e-3 --batch_size 32 --mode training --hyper_param_list hertz_8000 window_time_0.25 hop_time_0.125

  • python tuning_dl.py -m softmax -c trad -lr 1e-3 --batch_size 256 --mode training --hyper_param_list hertz_8000 window_time_0.25 hop_time_0.125 better than using batch size of 1024 and 512