๐น What problem is AComDim trying to solve?You have spectral data (NIR, Raman, etc.) collected under a designed experiment: ๐ข Several factors (e.g., instrument, temperature, sample type, operator)๐ข Possibly interactions…
Lees verder
UMAP: A Geometric Approach to Dimensionality Reduction
These six plots show how UMAP represents the structure of NIR milk spectra when coloured by fat percentage. UMAP builds a low-dimensional map by preserving neighbourhood relations from the original…
Lees verder๐ก The Crucial Role of Data Cleaning in Soil Spectroscopy Analysis
https://www.vibralytics.nl/wp-content/uploads/sites/617/2026/01/WhatsApp-Video-2025-12-08-at-13.51.37.mp4 In many data analysis projects, especially when dealing with complex soil spectra, the most time-consuming step is often not building sophisticated models, but rather thoroughly cleaning and preparing the…
Lees verder
Cleaning Extremely Noisy Soil Spectroscopy Data
Two-Criterion for Outlier Removal We have been diving deep into soil spectroscopy to predict key soil properties such as organic carbon, total carbon, nitrogen, clay, sand, silt, and bulk density.…
Lees verder
Unexpected Target Imbalance: What to do?
In NIR analysis, we have found spectral outliers, which we have removed using exploratory methods. However, we have also faced another challenge in this dataset: a severe imbalance in the…
Lees verderSNEEUW, SNEEUW, SNEEUW! Snow, nieve, what a beauty!
https://www.vibralytics.nl/wp-content/uploads/sites/617/2026/01/WhatsApp-Video-2026-01-09-at-10.35.37.mp4 We’ve been enjoying abundant snowfall here in the Netherlands, which reminds me so much of my stunning Patagonia… โ๏ธ What I find ABSOLUTELY FASCINATING about this research study, is…
Lees verder
Selection of NIR devices
Did you know itโs now possible to PRINT infrared detectors smaller than 10 micrometers? ๐ฎ๐ฎ In a recent study by Zhixuan Zhao and Ran An at the National University of…
Lees verder
๐ฌ๐ฒ๐ฟ๐ฏ๐ฎ ๐ ๐ฎ๐๐ฒ: ๐๐ผ๐ป๐ป๐ฒ๐ฐ๐๐ถ๐ป๐ด ๐ฃ๐ฒ๐ผ๐ฝ๐น๐ฒ ๐ณ๐ผ๐ฟ ๐ ๐ถ๐น๐น๐ฒ๐ป๐ป๐ถ๐ฎ ๐ฟโจ
Delicious, we canโt (and donโt want to) live without it. For many, it has been an integral part of daily life for thousands of years. ๐ฌ๐ฒ๐ฟ๐ฏ๐ฎ ๐บ๐ฎ๐๐ฒ (๐๐น๐ฒ๐ ๐ฝ๐ฎ๐ฟ๐ฎ๐ด๐๐ฎ๐ฟ๐ถ๐ฒ๐ป๐๐ถ๐) is…
Lees verder
โฝ Feature Selection in AI: Like Building a Winning Football Team
๐ Special thanks to Esmael Ahmed for making the dataset openly available on GitHub and Auckland Figshare:
๐ https://lnkd.in/eyHdPrrR
Revolutionizing Food Safety with AI: Visualizing Hyperparameter Optimization in NIR Spectroscopy!
This video showcases the optimization process for Artificial Neural Networks (ANN) and Convolutional Neural Networks (CNN) using Optuna, applied to NIR spectral data with various preprocessing treatments (e.g., RAW, SNV, Der1+SNV, Der2+SNV).
๐น What’s in the video?
X-Axis: Learning Rate (LR) โ Key to model convergence.
Y-Axis: Units/Filters in the first layer โ Representing network complexity.
Z-Axis: Accuracy โ Up to 1.0 in top models!
Colors: Blue for ANN, Red for CNN.
Animation: Rotates the 3D scatter plot to explore how hyperparameters evolve across treatments, revealing insights into which combos yield the best performance for detecting food adulteration.
This work highlights how AI can enhance chemometric analysis in food science, improving accuracy in identifying adulterants like those in oils, spices, or dairy products. CNN models often outperform ANN in spectral data, achieving near-perfect scores with optimal preprocessing!
If you’re in #NIRSpectroscopy, #AIFoodSafety, or battling #FoodAdulteration, let’s connect!