Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_plm_contiguousf4_20261105-043846

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
train3.63020.42670.45760.14802870.00000.22016656.00000.80001.62874.16940.00000.00000.39240.72001.35975.34230.02000.00000.41540.66001.07196.15300.07000.00000.40700.62571.08196.31790.12870.00000.4383
validation2.17620.23190.16780.04211218.00000.017680.00000.4000-0.47461.87170.00000.00000.06590.3800-0.39641.87170.00000.00000.10030.3200-0.44485.64210.02000.00000.10220.3256-0.51971.87170.00000.00000.0689
test2.39310.38880.33000.1406933.0000-0.05821138.00000.1000-1.34030.30150.00000.0000-0.04600.2600-0.87101.54560.00000.00000.05450.2600-0.73841.54560.00000.00000.10060.2368-1.01710.45660.00000.00000.0182

Loss per epoch

training validation

Config

name: reference_ube4b_fc_plm_contiguousf4
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index,plm,plm_sum
  split_dir: null
  split_scheme: contiguous
  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