Protein Engineering

gb1 · Fully connected

reference_gb1_fc_plm_contiguousf3_20261105-040600

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
train0.42220.97120.97570.913430012.00000.8123130.00001.00001.67911.85270.00000.00000.88971.00001.64842.00800.18000.00000.93291.00001.64472.01760.30000.00000.95341.00001.32972.52950.75510.00000.9823
validation5.10190.14890.0452-0.15045856.00000.012711966.00000.2000-1.05930.29380.00000.00000.17320.1000-1.34830.29380.00000.00000.07700.1100-1.32190.76080.00000.00000.08850.1095-1.29890.76080.00000.00000.1002
test6.87640.28950.2786-0.25382473.0000-0.02186563.00000.3000-2.62740.81340.00000.00000.15890.4600-1.27871.02550.00000.00000.43640.4300-1.50551.02550.00000.00000.39560.4750-1.34861.40540.01000.00000.4366

Loss per epoch

training validation

Config

name: reference_gb1_fc_plm_contiguousf3
task: dms
data:
  dataset: gb1
  encoding: one_hot,aa_index,plm,plm_sum
  split_dir: null
  split_scheme: contiguous
  split_fold: 3
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