Correcting Differences with Machine Learning

Δ-ML approach drastically reduces mean absolute error (MAE) in atomization enthalpies for different QC methods

Δ-ML approach drastically reduces mean absolute error (MAE) in atomization enthalpies for PM7 and B3LYP methods in respect to accurate G4MP2 values

In our recent study, we propose using machine learning (ML) to correct differences in properties calculated at two quantum chemical (QC) methods with different accuracy.

In the Δ-ML approach ML model is trained on differences between some property \(P_t\) calculated at the target level of theory and property \(P’_b\) calculated at the baseline level of theory. Then this ML model is used to predict \(\Delta_b^t\) for out-of-sample molecules.

Read more ›

Tagged with: , , , , , ,

Machine Learning of Semiempirical Parameters

We propose using machine learning (ML) for improving semiempirical Hamiltonian. Given sufficiently large training set ML can be used to correct parameters of semiempirical quantum chemical (SQC) method individually for any target molecule.

Improvement of a semiempirical quantum chemical (SQC) method using Machine Learning (ML)

Improvement of a semiempirical quantum chemical (SQC) method using Machine Learning (ML)

Such automatic parametrization technique (APT) stands in stark contrast to the traditional special-purpose reparametrization (SPR), when parameters are optimized for specific type of molecules and then resulting rSQC method is used unchanged for every other target molecule.

For our studies we used subset of huge database published by us. Hybrid ML-SQC approach has much lower error in predicted atomization enthalpies in comparison with SQC method with standard parameters.

Read more ›

Tagged with: , , , ,

National Wealth and High-Quality Research

In a response to my previous post “Geography of High-Quality Science” Rasmus Persson has raised an interesting question: to what extent does or does not national GDP correlate with the “research output”? Here I try to answer it and analyze trends.

To plot GDP values vs respective weighted fractional counts (WFCs) for each country covered by Nature Index, I have taken the same WFCs as in my previous post and GDP estimates derived from purchasing power parity (PPP) calculations as listed in the CIA World Factbook (in billions of international dollars, data retrieved from the Wikipedia article, the GDP for Vatican is derived from National Geographic page). Keep in mind that GDP values are different in different sources, but the general trends must be similar.

Read more ›

Tagged with: , ,

Geography of High-Quality Science

Nature publishing group introduced Nature Index at the end of 2014. This index is essentially a collection of author affiliations mentioned in the research papers published in 68 selected high-quality journals. Nature index can be used (and respective tools are provided in NPG website) to assess the high-quality scientific output by region, research institution and subject. Obviously Index by itself does not say anything about the quality level of an individual study or researcher. Nevertheless, it is quite useful to estimate general trends.

Read more ›

Tagged with: ,

New 1D Coordination Polymers


New paddlewheel one-dimensional coordination complexes, M = Rh, Ru and Mo

In our study we reported synthesis, and experimental and theoretical characterization of new one-dimensional coordination polymers. Research article “Multiply Bonded Metal(II) Acetate (Rhodium, Ruthenium, and Molybdenum) Complexes with the trans-1,2-Bis(N-methylimidazol-2-yl)ethylene Ligand” (DOI: 10.1021/ic501435a) was published on November 13th in the Inorganic Chemistry. The experimental work was done in the group of Prof. Dr. Nicolai Burzlaff and theoretical part in the group of Prof. Dr. Timothy Clark.

Read more ›

Tagged with:

Data Set with 134 Kilo Molecules

If you need really huge data set to test your methods, then our data set with 133,885 species is one of the best choices. You can download it in figshare.

Distribution of species according to number of electron pairs, Figure from the data descriptor.

Read more ›

Tagged with: , , , , ,

The Unrestricted Local Properties

Did you know that the reactivity of alkyl radicals towards H-abstraction is related to their electron accepting properties? And that alkyl cations are much more reactive than alkyl radicals for the same reason? The same tool that clearly visualizes these effects helps to explain what channels are responsible for electron and hole transport in ambipolar transistor. Find out more in my recent paper.

Tagged with: , , , ,