Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_modulof4_20261104-170252

Metrics

Setmsepearsonrspearmanrr2above_wt_nabove_wt_spearmanrbudget_to_maxhit_rate_at_10mean_at_10max_at_10precision_at_10tie_share_at_10captured_at_10hit_rate_at_50mean_at_50max_at_50precision_at_50tie_share_at_50captured_at_50hit_rate_at_100mean_at_100max_at_100precision_at_100tie_share_at_100captured_at_100hit_rate_at_1pctmean_at_1pctmax_at_1pctprecision_at_1pcttie_share_at_1pctcaptured_at_1pct
train2.49650.61050.57320.35616783.00000.478339.00000.90004.56297.82590.10000.00000.66730.92003.96189.15370.24000.00000.70340.93003.71389.15370.29000.00000.72490.94043.48049.15370.46360.00000.7904
validation3.3086-0.0595-0.0683-0.0139645.00000.0262504.00000.4000-0.94573.41600.00000.00000.04950.2000-1.57463.41600.02000.0000-0.07630.1800-1.56583.41600.04000.0000-0.08500.2500-1.54493.41600.03570.0000-0.0638
test3.7114-0.0722-0.0773-0.0723318.00000.01761626.00000.2000-1.61981.59700.00000.0000-0.00250.0800-2.41561.59700.02000.0000-0.19560.0700-2.31301.59700.02000.0000-0.20110.1111-2.33091.59700.00000.0000-0.1435

Loss per epoch

training validation

Config

name: reference_ube4b_fc_modulof4
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: modulo
  split_fold: 4
model:
  name: fcn
  hidden:
  - 100
  - 100
  - 100
  dropout: 0.2
train:
  epochs: 300
  batch_size: 64
  learning_rate: 0.0001
  patience: 15
  min_delta: 1.0e-05
  seed: 7