Reproduce “Machine Learning for Raman Spectroscopy-based Cyber-Marine Fish Biochemical Composition Analysis” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yun Zhou, Gang Chen, Bing Xue, Mengjie Zhang, Jeremy S. Rooney, Kirill Lagutin, Andrew MacKenzie, Keith C. Gordon and 1 more
Researchers developed FishCNN, a Convolutional Neural Network trained on Raman spectroscopic data, to enable rapid, non-destructive assessment of water, protein, and lipid content in fish. The model achieved an R² value of approximately 0.81 for multi-output regression, outperforming traditional methods and other CNNs on small datasets through tailored preprocessing and data augmentation.
Key insights
- The order of operations in data preparation is critical, with data augmentation applied after preprocessing yielding better model performance.
- A specialized CNN architecture with large convolutional kernels effectively captures broad patterns in one-dimensional spectral data, suitable for small datasets.
- InGaAs Raman spectroscopy outperformed FT-Raman for biological samples due to reduced fluorescence interference, despite its narrower spectral range.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.