Open
Conversation
Implements the two-stage DILA (Dictionary Label Attention) pipeline: - SparseAutoencoder for learning sparse dictionary features from PLM embeddings - DictionaryLabelAttention for ICD-guided label attention mechanism - DILA BaseModel integrating both stages with pretrain_sparse_autoencoder utility - Unit and integration tests covering all three modules Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
dila model and ablation script
Add embeddings and sparse autoencoder implementation
feat: add DILA model for interpretable ICD coding
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Contributors
Type of Contribution
Link to Original Paper
High-level Description
This PR implements the Dictionary Label Attention (DILA) model for multi-label clinical classification tasks, such as ICD coding. DILA provides mechanistic interpretability by disentangling dense embeddings into sparse, human-auditable medical concepts.
File Guide
docs/api/models.rstdocs/api/models/pyhealth.models.dila.rst: API documentation for the model.pyhealth/models/dila.py: Main model class integrating PLM, SAE, and Attention modules.pyhealth/models/dila_sparse_autoencoder.py: SAE implementation with centering and elastic-net loss.pyhealth/models/dila_dict_label_attention.py: Label attention logic and ICD projection initialization.tests/core/test_dila.py: Fast unit tests using synthetic tensors.examples/mimic3_icd_coding_dila.py: Script for hyperparameter ablation studies.examples/dila_mimic3_evaluation.ipynb: Notebook for visualization and evaluation.