[CS598] Add MultiViewContrastive for medical time-series domain adaptation (Oh & Bui, CHIL 2025)

ablation_results.json

@@ -0,0 +1,100 @@
+{
+  "main_reproduction": {
+    "accuracy": 0.8732,
+    "accuracy_std": 0.031371,
+    "precision_macro": 0.890223,
+    "precision_std": 0.022084,
+    "recall_macro": 0.8732,
+    "recall_std": 0.031371,
+    "f1_macro": 0.871541,
+    "f1_std": 0.032488,
+    "n_seeds": 10,
+    "pretrain_time_s": 5038.404446363449,
+    "finetune_time_s": 7.1890788078308105,
+    "n_params": 468546
+  },
+  "encoder_transformer": {
+    "accuracy": 0.8732,
+    "accuracy_std": 0.031371,
+    "precision_macro": 0.890223,
+    "precision_std": 0.022084,
+    "recall_macro": 0.8732,
+    "recall_std": 0.031371,
+    "f1_macro": 0.871541,
+    "f1_std": 0.032488,
+    "n_seeds": 10,
+    "pretrain_time_s": 5038.404446363449,
+    "finetune_time_s": 7.1890788078308105,
+    "n_params": 468546
+  },
+  "fusion_attention": {
+    "accuracy": 0.8732,
+    "accuracy_std": 0.031371,
+    "precision_macro": 0.890223,
+    "precision_std": 0.022084,
+    "recall_macro": 0.8732,
+    "recall_std": 0.031371,
+    "f1_macro": 0.871541,
+    "f1_std": 0.032488,
+    "n_seeds": 10,
+    "pretrain_time_s": 5038.404446363449,
+    "finetune_time_s": 7.1890788078308105,
+    "n_params": 468546
+  },
+  "encoder_cnn": {
+    "accuracy": 0.9224,
+    "accuracy_std": 0.022553,
+    "precision_macro": 0.927741,
+    "precision_std": 0.015151,
+    "recall_macro": 0.9224,
+    "recall_std": 0.022553,
+    "f1_macro": 0.922029,
+    "f1_std": 0.023281,
+    "n_seeds": 10,
+    "pretrain_time_s": 835.3022849559784,
+    "finetune_time_s": 3.418339490890503,
+    "n_params": 392130
+  },
+  "encoder_gru": {
+    "accuracy": 0.7704,
+    "accuracy_std": 0.085971,
+    "precision_macro": 0.841265,
+    "precision_std": 0.042425,
+    "recall_macro": 0.7704,
+    "recall_std": 0.085971,
+    "f1_macro": 0.752609,
+    "f1_std": 0.100316,
+    "n_seeds": 10,
+    "pretrain_time_s": 1774.1677017211914,
+    "finetune_time_s": 3.1605217456817627,
+    "n_params": 355266
+  },
+  "fusion_concat": {
+    "accuracy": 0.839,
+    "accuracy_std": 0.044247,
+    "precision_macro": 0.868079,
+    "precision_std": 0.025843,
+    "recall_macro": 0.839,
+    "recall_std": 0.044247,
+    "f1_macro": 0.834898,
+    "f1_std": 0.047727,
+    "n_seeds": 10,
+    "pretrain_time_s": 5038.404446363449,
+    "finetune_time_s": 7.5263755321502686,
+    "n_params": 419010
+  },
+  "fusion_mean": {
+    "accuracy": 0.8658,
+    "accuracy_std": 0.025416,
+    "precision_macro": 0.885845,
+    "precision_std": 0.014511,
+    "recall_macro": 0.8658,
+    "recall_std": 0.025416,
+    "f1_macro": 0.863777,
+    "f1_std": 0.027056,
+    "n_seeds": 10,
+    "pretrain_time_s": 5038.404446363449,
+    "finetune_time_s": 7.912641763687134,
+    "n_params": 418498
+  }
+}

docs/api/models.rst

@@ -187,6 +187,7 @@ API Reference
     models/pyhealth.models.EHRMamba
     models/pyhealth.models.JambaEHR
     models/pyhealth.models.ContraWR
+    models/pyhealth.models.MultiViewContrastive
     models/pyhealth.models.SparcNet
     models/pyhealth.models.StageNet
     models/pyhealth.models.StageAttentionNet

docs/api/models/pyhealth.models.MultiViewContrastive.rst

@@ -0,0 +1,14 @@
+pyhealth.models.MultiViewContrastive
+======================================
+
+Multi-View Contrastive Learning for Domain Adaptation in Medical Time Series.
+
+Paper: Oh, Y.; and Bui, A. 2025. Multi-View Contrastive Learning for Robust
+Domain Adaptation in Medical Time Series Analysis. In *Proceedings of the
+Sixth Conference on Health, Inference, and Learning*, volume 287, 502--526.
+PMLR.
+
+.. autoclass:: pyhealth.models.MultiViewContrastive
+    :members:
+    :undoc-members:
+    :show-inheritance: