We compared different appraoches for optimisation of multitask neural network hyperparameters on QSAR data, results were recently published in the following manuscript: Valsecchi, C., Consonni, V., Todeschini, R., Orlandi, M., Gosetti, F., Ballabio, D. (2021) Parsimonious Optimization of Multitask Neural Network Hyperparameters, Molecules, 26, 7254 Have a look here!
News
Welcome to Veronica Termopoli
Welcome to Veronica Termopoli, new member of the Milano Chemometrics and QSAR Research Group! Veronica has a strong background on chromatography and she will give her introductory seminar “New instrumental and applicative approaches in LC-MS” at our department next wednesday november 10th, more info here.
Classification toolbox: new release!
A new release of the classification toolbox for MATLAB is now available: variance scaling and range scaling have been added as approaches for column pre-processing. Basic row pre-processing methods have been added (standard normal variate, multiplicatve scatter correction, first and second derivative). Savitzky-Golay smoothing can now be applied to analytical spectral data. The layout of the calculation menu has been simplified. Download the latest version here!
Kohonen and CPANN Toolbox: new release out!
A new release of the Kohonen and CPANN Toolbox (for Matlab) is now available! The toolbox is a collection of MATLAB modules for calculating Kohonen Maps and Counterpropagation Artificial Neural networs (CPANNs), Supervised Kohonen networks and XY-fused networks. In this new release, it is possible to have calculation on a GPU. This can be helpfull when dealing with big data. The toolbox can be downloaded here.
Michem video to promote degrees in Chemical Sciences @ UNIMIB
We recently participated to the action promoted by the Degree in Chemical Sciences at the University of Milano – Bicocca to promote the courses around Italy. Here the result (in Italian…): https://www.youtube.com/watch?v=ClRF7sf0hxU
Kohonen and CPANN Toolbox: new release!
A new release of the Kohonen and CPANN Toolbox (for Matlab) is now available! The toolbox is a collection of MATLAB modules for calculating Kohonen Maps and Counterpropagation Artificial Neural networs (CPANNs), Supervised Kohonen networks and XY-fused networks. In this new release, the computational time for the calculation of SOMs has been significantly reduced. The toolbox can be downloaded here.
Regression toolbox for MATLAB now available!
   |
The Regression toolbox (for MATLAB) is a collection of MATLAB modules for calculating regression multivariate models: Ordinary Least Squares (OLS), Partial Least Squares (PLS), Principal Component Regression (PCR), Ridge regression, local regression based on K Nearest Neighbours (KNN) and Binned Nearest Neighbours (BNN) approaches, and variable selection approaches (All Subset Models, Forward selection, Genetic Algorithms and Reshaped Sequential Replacement).
The toolbox is freely available and can be downloaded here |
New publication on Collaborative Acute Toxicity Modeling
The manuscript about our latest partecipation in the CATMoS collaborative modelling project to predict Acute Oral Toxicity is not out, have a look: CATMoS: Collaborative Acute Toxicity Modeling Suite, Environmental Health Perspectives, 129, 47013 [link]
Special issue on Separations
Special Issue “UHPLC-MS/MS Methods for the Identification of Emerging Contaminant Transformation Products in Surface Water”, editor: Fabio Gosetti, have a look here if interested: https://www.mdpi.com/journal/separations/special_issues/UHPLC_Water
New master thesis in chemometrics!
We welcome Leonardo Fedrigotti, Martina Barbagallo and Davide crucitti, who will start their master thesis on chemometrics related topics. In particular, Martina will develop and validate proper chemometrics strategies based on advanced neural networks and deep-learning models to predict the molecular structure of some substances starting from their LC-MS/MS spectra, which were previously organized in a database. Davide will work in the field of QSAR, developing a model to classify substrates, inhibitors and non-active compounds of P-Glycoprotein using artificial neural networks. His project will be carried out at the Slovenian National Institute of Chemistry. Leonardo will develop a model which is able to couple GC-MS data and features of encephalographic responses (EEG) measured with portable low-cost devices to assess odor-stimulated emotions and to ensure scientific measurements of such signals. Coupling and comparison of analytical sources will be carried out through chemometrics methods. This project will be carried out at P&G (Bruxelles).