Shift invariant soft trilinearity: Modelling shifts and shape changes in gas-chromatography coupled mass spectrometry

• The proposed method is a shift-invariant tri-linear decomposition that can effectively model peak shape changes in gas chromatography coupled mass spectrometry (GC-MS). The method automatically adapts to the complexity of the data, allowing for a fine-tuned flexibility in the modeling process. Performance on real GC-MS data indicates that the proposed method extends the shift-invariant tri-linearity constraint (SIT) and improves factor resolution if the data does not behave tri-linearly. Although the SIST method is more flexible than SIT, experiments with real data indicate that it suffers significantly less from rotational ambiguity than non-negative MCR-ALS. Furthermore, results on real GC-MS data indicate that SIST has faster convergence than PARAFAC2 and suffers less from local minima.