Home

Team 09 / Model Distillation

Organization

• Weekly meeting every Monday 11 am. Zoom-Link

Workflow

• Master branch is protected
• Merge to master by create pull request and approval by team members
• SCRUM like process. Use 'Projects' tab
  • Create Issue as a ticket
  • ONLY work on this issue, if another problem arises, open another issue.
  • Create branch from issue
  • If work is fulfilled, create pull request
• Codestyle: Type Hints.
• Docstyle: As described here.

Papers

ML-Papers

• Model Distillation: A Survey
• Positive-Unlabeled Learning with Non-Negative Risk Estimator
• Positive-Unlabeled Learning from Imbalanced Data

Background Knowledge

• Joint epitope selection and spacer design for string-of-beads vaccines
• Integrated modeling of the major events in the MHC class I antigen processing pathway
• The role of the proteasome in generating cytotoxic T-cell epitopes: insights obtained from improved predictions of proteasomal cleavage
• Bounding and Counting Linear Regions of Deep Neural Networks (not part of this project)

Useful Links

• Summary of Model Distillation Survey

Code Inspiration

• [RCNN Distillation] (https://github.com/DoctorKey/LSHFM.detection)
• [Distillation in Keras] (https://keras.io/examples/vision/knowledge_distillation/)
• [Knowledge Distillation of DNNs] (https://github.com/peterliht/knowledge-distillation-pytorch)
• [Promising Implementation] (https://github.com/AberHu/Knowledge-Distillation-Zoo/blob/master/train_kd.py)