Transfer Learning for Raman Spectroscopy: My Approach to the DIG-4-BIO Challenge
In this post, I walk through my proposal for the DIG-4-BIO Raman Transfer Learning Challenge, where I achieved a public R² score of 0.7491 and reached 3rd place (as of writing). The task involves predicting chemical concentrations from Raman spectra collected from various instruments—a highly realistic and noisy setting. I detail how I cleaned, augmented, modelled, and evaluated this complex dataset.
Aug 1, 2025
