Closed richelbilderbeek closed 2 years ago
Due to the gcaer-only workflow, this will be very easy.
[richel@rackham2 ~]$ cat 20_start_issue_18.log
Starting time: 2022-04-08T10:41:07+0200
Running on computer with HOSTNAME: r44
Running at location /home/richel
gcae_experiment_params_filename: /home/richel/sim_data_issue_18/experiment_params.csv
unique_id: issue_18
sbatch: error: Invalid directive found in batch script: defined
End time: 2022-04-08T10:41:07+0200
Duration: 0 seconds
[richel@rackham2 ~]$
[richel@rackham2 ~]$ ./nsphs_ml_qt/scripts_rackham/20_start_issue_18.sh
Starting time: 2022-04-08T11:27:36+0200
Running on computer with HOSTNAME: rackham2.uppmax.uu.se
Running at location /home/richel
gcae_experiment_params_filename: /home/richel/sim_data_issue_18/experiment_params.csv
unique_id: issue_18
jobid_21: 25760509
jobid_22: 25760510
jobid_25: 25760511
End time: 2022-04-08T11:27:36+0200
Duration: 0 seconds
[richel@rackham2 ~]$ q
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
25760509 core 21_creat richel PD 0:00 1 (None)
25760510 core 22_creat richel PD 0:00 1 (None)
25760511 core 25_run.s richel PD 0:00 1 (Dependency)
[richel@rackham2 ~]$ q
We are not interested in this anymore, for now. First use real data, so the autoencoder is not underwhelmed, then take it from there.
Done!
Depends on
Currently, quantitative trait prediction is poor for simple settings. This is probably due to the latent layer consisting of
75 neurons, which is overkill2 neurons.Run the autoencode on simple settings with a latent layer of 1..10 neurons,