Mohammed-Yassine Habibi

Mohammed-Yassine Habibi

PhD student in interpretable machine learning & concept learning for ecology · Inria (Thoth) & LECA · Corps des IPEF

Kernel Methods for DNA Sequence Classification

Winter 2025 · MVA — Kernel Methods for Machine Learning

Code

  • Kernel methods
  • DNA sequences

Kaggle data challenge of the MVA course Kernel Methods for Machine Learning: binary classification of DNA sequences on three datasets, with kernel methods implemented from scratch.

  • String kernels implemented from scratch: spectrum, mismatch and gapped k-mer kernels (plus Gaussian kernels on numerical embeddings);
  • Kernel combination (products, sums and weights chosen by kernel–target alignment);
  • Kernel SVM, with the quadratic program solved via OSQP, and cross-validated model selection.