Paul W. Ayers and Farnaz Heidar-Zadeh
Can machine-learning methods reveal chemical rules-of-thumb? Consider, for example, the concept of acidity. If A is a stronger Lewis acid than B, then A will accept electrons from B. Similarly, if B is a stronger Lewis acid than C, then B will accept electrons from C. The concept of acid/base strength indicates that A will also accept electrons from C. (Of course, there are numerous exceptions to this rule, which chemists excuse by asserting that acid/base strength is not always the dominant effect in a chemical reaction.) By viewing acid/base reactions as a “competition for electrons,” we deduce the acid/base strength of different reagents. This provides an alternative scale for Lewis acidity. This study also presents a template for how machine-learning ideas can be used to deduce chemical concepts and rules-of-thumb.
Back to Machine Learning for Many-Particle Systems