Protein Engineering

ube4b · Fully connected

reference_ube4b_fc_plm_randomf1_20261104-180738

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
train1.98110.69120.65620.477012540.00000.460097.00001.00005.55247.74370.20000.00000.76981.00005.29848.62460.28000.00000.84981.00005.17749.15370.39000.00000.88381.00003.98889.15370.57970.00000.8971
validation2.14110.66330.62200.43964076.00000.44701.00001.00005.67109.18450.20000.00000.79981.00004.80099.18450.38000.00000.85250.99004.41569.18450.45000.00000.86930.98983.99339.18450.59390.00000.8849
test2.10970.66630.62430.44314135.00000.4476201.00001.00005.15146.57580.10000.00000.81181.00004.31616.57580.34000.00000.80361.00004.29736.57580.49000.00000.86680.98983.84996.57580.52790.00000.8730

Loss per epoch

training validation

Config

name: reference_ube4b_fc_plm_randomf1
task: dms
data:
  dataset: ube4b
  encoding: one_hot,aa_index,plm,plm_sum
  split_dir: null
  split_scheme: random
  split_fold: 1
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