Reports
Created by
Created On
Last edited
CNEP Training Frozen ST+LSTMCNN
Training of frozen Encoders with preloaded parameters, Embedding dense layers were not frozen. Training was done with 5 different learning rates.
1
thetaphipsi2021-12-11
Training CNEP with different LR & Warmup
Setup:
1. Sentence-Transformer pretrained as encoder for MIMIC-III Clinical Notes,
2. LSTMCNN + Dense Layer pretrained on MIMIC-III/Readmission (w/o Clinical Notes) as encoder for Events and charts data
0
2021-12-12
Training CNEP with different LR & Warmup
Setup:
1. Sentence-Transformer pretrained as encoder for MIMIC-III Clinical Notes,
2. LSTMCNN + Dense Layer pretrained on MIMIC-III/Readmission (w/o Clinical Notes) as encoder for Events and charts data
1
2021-12-12