train all models with pmc article/det compressed embeddings

[source-data]

No description provided.

[source-data]

Model: 5X_L1200_anonym_not_reporter_article_embeddings_128_intervention_assayed_2019-08-22-16-25.zip

[source-data]

Benchmarking results:

========================================================

Data: /workspace/py-smtag/resources/data4th/5X_L1200_article_embeddings_128

Model: 5X_L1200_article_embeddings_128_small_molecule_geneprod_subcellular_cell_tissue_organism_assay_2019-08-23-17-46.zip

Global stats:

precision = 0.8090225458145142                                                                                                                          
recall = 0.78                                                                                                                                           
f1 = 0.79                                                                                                                                               

Feature: 'small_molecule'

precision = 0.72
recall = 0.72
f1 = 0.72

Feature: 'geneprod'

precision = 0.84
recall = 0.88
f1 = 0.86

Feature: 'subcellular'

precision = 0.72
recall = 0.78
f1 = 0.75

Feature: 'cell'

precision = 0.83
recall = 0.78
f1 = 0.81

Feature: 'tissue'

precision = 0.77
recall = 0.67
f1 = 0.72

Feature: 'organism'

precision = 0.83
recall = 0.67
f1 = 0.74

Feature: 'assay'

precision = 0.77
recall = 0.70
f1 = 0.73

Feature: 'untagged'

precision = 0.99
recall = 0.99
f1 = 0.99

========================================================

Data: /workspace/py-smtag/resources/data4th/5X_L1200_anonym_not_reporter_article_embeddings_128

Model: 5X_L1200_anonym_not_reporter_article_embeddings_128_intervention_assayed_2019-08-22-16-25.zip

Global stats:

precision = 0.8826258778572083
recall = 0.88
f1 = 0.88

Feature: 'intervention'

precision = 0.82
recall = 0.76
f1 = 0.79

Feature: 'assayed'

precision = 0.83
recall = 0.87
f1 = 0.85

Feature: 'untagged'

precision = 1.00
recall = 1.00
f1 = 1.00

========================================================

Data: /workspace/py-smtag/resources/data4th/5X_L1200_molecule_anonym_article_embeddings_128

Model: 5X_L1200_molecule_anonym_article_embeddings_128_intervention_assayed_2019-08-28-23-33_epoch_51.zip

Global stats:

precision = 0.8728659749031067
recall = 0.90
f1 = 0.88

Feature: 'intervention'

precision = 0.95
recall = 0.92
f1 = 0.94

Feature: 'assayed'

precision = 0.67
recall = 0.78
f1 = 0.72

Feature: 'untagged'

precision = 1.00
recall = 1.00
f1 = 1.00

========================================================

Data: /workspace/py-smtag/resources/data4th/5X_L1200_article_embeddings_128

Model: 5X_L1200_article_embeddings_128_reporter_2019-08-28-00-08_epoch_23_.zip

Global stats:

precision = 0.9299678206443787
recall = 0.88
f1 = 0.90

Feature: 'reporter'

precision = 0.86
recall = 0.76
f1 = 0.81

Feature: 'untagged'

precision = 1.00
recall = 1.00
f1 = 1.00

========================================================

Data: /workspace/py-smtag/resources/data4th/5X_L1200_emboj_2012_no_viz

Model: 5X_L1200_emboj_2012_no_viz_panel_stop_2019-08-29-08-31.zip

Global stats:

precision = 0.9664154648780823
recall = 1.00
f1 = 0.98

Feature: 'panel_stop'

precision = 0.93
recall = 0.99
f1 = 0.96

Feature: 'untagged'

precision = 1.00
recall = 1.00
f1 = 1.00

========================================================

Data: /workspace/py-smtag/resources/data4th/10X_L1200_disease_articke_embeddings_128 + /workspace/py-smtag/resources/data4th/5X_L1200_article_embeddings_128

Model: 10X_L1200_disease_articke_embeddings_128-5X_L1200_article_embeddings_128_disease_2019-08-25-21-47.zip

Global stats:

precision = 0.9544399380683899
recall = 0.89
f1 = 0.92

Feature: 'disease'

precision = 0.91
recall = 0.79
f1 = 0.85

Feature: 'untagged'

precision = 1.00
recall = 1.00
f1 = 1.00