Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_randomf4_20261104-165127

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
train1.97020.69530.65980.481112414.00000.46412.00001.00006.49889.18450.20000.00000.86881.00005.37789.18450.30000.00000.85191.00005.08849.18450.42000.00000.87091.00004.05799.18450.58640.00000.9063
validation2.18150.64560.60760.41654174.00000.424632.00001.00005.04097.82590.20000.00000.76381.00004.62268.62460.38000.00000.83670.99004.39978.62460.51000.00000.87990.97973.76278.62460.55840.00000.8584
test2.19760.65490.61780.42854163.00000.464349.00001.00005.54997.74370.30000.00000.83591.00004.73569.15370.34000.00000.85661.00004.35759.15370.46000.00000.87010.98483.83559.15370.54820.00000.8655

Loss per epoch

training validation

Config

name: reference_ube4b_fc_randomf4
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: random
  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