Protein Engineering

gb1 · Fully connected

reference_gb1_fc_randomf4_20261104-163141

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
train0.19950.98070.97970.961656646.00000.864693.00001.00001.65691.92400.00000.00000.87831.00001.66382.19520.10000.00000.92651.00001.68112.52950.29000.00000.95190.99971.30302.52950.79610.00000.9875
validation0.21640.97900.97780.958418587.00000.8552515.00001.00001.71541.99510.10000.00000.92611.00001.65132.09750.30000.00000.95451.00001.61652.09750.46000.00000.96681.00001.29102.45390.78730.00000.9858
test0.21300.97930.97830.959018803.00000.8670135.00001.00001.66251.93620.10000.00000.91641.00001.58552.08880.28000.00000.94041.00001.57902.08880.40000.00000.95951.00001.31502.14550.80970.00000.9882

Loss per epoch

training validation

Config

name: reference_gb1_fc_randomf4
task: dms
data:
  dataset: gb1
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: random
  split_fold: 4
model:
  name: fcn
  hidden:
  - 1000
  dropout: 0.2
train:
  epochs: 300
  batch_size: 64
  learning_rate: 0.0001
  patience: 15
  min_delta: 1.0e-05
  seed: 7