Protein Engineering

gb1 · Fully connected

reference_gb1_fc_modulof4_20261104-165808

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
train0.38390.96320.96890.924735489.00000.8077166.00001.00001.73621.99510.00000.00000.89961.00001.67302.11100.14000.00000.93071.00001.63612.11100.25000.00000.94171.00001.36662.52950.76080.00000.9825
validation6.88780.00510.0197-0.23402689.0000-0.009711325.00000.2000-3.92400.49260.00000.0000-0.28700.2200-2.97220.49260.00000.0000-0.05290.1800-2.75630.78470.00000.00000.00380.2239-2.39520.78470.00500.00000.1033
test4.91630.01640.0192-0.06403230.0000-0.09043732.00000.0000-1.3775-0.35600.00000.00000.23480.1000-1.36720.26250.00000.00000.24670.1200-1.70370.26250.00000.00000.15850.1592-1.73351.12060.00500.00000.1543

Loss per epoch

training validation

Config

name: reference_gb1_fc_modulof4
task: dms
data:
  dataset: gb1
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: modulo
  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