🤖 AI Summary
This study addresses the prevalent use of generic convolutional neural network (CNN) architectures in near-infrared (NIR) spectroscopic chemometrics, which often overlooks critical data-specific characteristics such as sampling properties, smoothness, redundancy, and sample size. By systematically analyzing 25 NIR regression tasks, the authors develop an interpretable one-dimensional CNN backbone and integrate spectral descriptors—including spectral entropy, intrinsic rank, autocorrelation, and wavelet scales—to establish, for the first time, a mapping between spectral properties and optimal CNN hyperparameters. This mapping is derived via Bayesian hyperparameter optimization and leave-one-dataset-out validation. The resulting transferable warm-start heuristic significantly reduces reliance on costly hyperparameter searches, achieving median test RMSE ratios of 0.953 and 1.017 under direct and leave-one-out evaluation, respectively—performance comparable to fully optimized models with equivalent sensitivity to random seeds.
📝 Abstract
Convolutional neural networks (CNN) for near-infrared (NIR) chemometrics are often designed using generic architectural rules, although spectral datasets differ in sampling, smoothness, redundancy, and sample size. We tested whether these properties can provide empirical priors for CNN design. Across 25 NIR regression tasks, we computed descriptors of dataset size, spectral length and spacing, entropy, intrinsic rank, autocorrelation, and wavelet-scale structure. Two interpretable 1D-CNN scaffolds (a minimal single-convolution model and an extended shallow model with optional branching, dilation, etc) were optimized using five-fold cross-validated Bayesian hyperparameter optimization (HPO). Relationships extracted from near-optimal trials were converted into warm-start heuristics and evaluated directly and through leave-one-dataset-out (LODO) validation. The clearest relationships involved convolutional receptive fields. In the minimal CNN, the preferred kernel fraction decreased with spectral entropy and intrinsic rank, increased with the wavelet energy-support fraction, and the learning rate tended to decrease with training-set size. Direct and LODO heuristics were competitive with HPO, with median test-RMSE ratios of 0.953 and 1.017, respectively. The extended CNN showed similar but less transferable structure across branch usage, dilation, dropout, filter counts, and receptive-field choices. Ten stochastic refits showed seed sensitivity comparable to that of HPO-selected configurations. In a separate experiment, joint preprocessing and CNN HPO outperformed standardized-spectra HPO in 19 of 25 tasks, although gains were dataset-dependent. These results show that spectral descriptors can provide practical CNN design priors, guiding shallow NIR models toward plausible hyperparameter regions before target-specific tuning