Adsorption energies of surface species are critical descriptors for catalytic materials design . The need to produce accurate prediction of descriptors has been ever increasing in recent years. Quantum mechanics provides a well-established framework to do this, notably through density-functional theory calculations. Exploration of large materials space through DFT calculated descriptors is really time-consuming and is out of reach for current computing facility. This is one of the reasons for the continuing efforts in chemisorption models, such as the d-band model [2,3], which provide fast access to adsorption energies based on properties of clean surfaces. Such simple models are extremely useful to understand trends of surface chemisorption. However, for many complex chemical situations accurate models do not yet exist. More recently approaches based on machine learning using neural networks or Gaussian processes (GPs) to fit DFT-calculated total energies using atomic positions/types as input features requires too many calculations to be feasible for catalyst screening purpose [4,5]. Here, we propose an alternative machine-learning (ML) based scheme where ab-initio adsorption energies and electronic fingerprint of clean metal surfaces are used to augment the ML database, enabling accurate prediction. The scheme could equally be viewed as an improved chemisorption model where complex is built in by non-linearity of the neural network. The scheme can be integrated with linear scaling relationships [6,7] established in surface chemistry for accelerating discovery of new catalysts. We used biomass-derived polyols such as ethylene glycol for fuels and chemicals as test reactions. We illustrated that our approach can greatly reduce computational time required for high-through screening and identify multimetallic alloys with improved efficiency and selectivity.
 J. K. Nørskov, F. Abild-Pedersen, F. Studt, and T. Bligaard, Proc. Natl. Acad. Sci. 108, 937 (2011).
 B. Hammer and J. K. Nørskov, Surf. Sci. 343, 211 (1995).
 H. Xin, A. Holewinski, and S. Linic, ACS Catal. 2, 12 (2012).
 J. Behler and M. Parrinello, Phys. Rev. Lett. 98, 146401 (2007).
 A. P. Bartók, M. C. Payne, R. Kondor, and G. Csányi, Phys. Rev. Lett. 104, 136403 (2010).
 F. Abild-Pedersen, J. Greeley, F. Studt, J. Rossmeisl, T. R. Munter, P. G. Moses, E. Skulason, T. Bligaard, and J. K. Norskov, Phys. Rev. Lett. 99, 016105 (2007).
 T. Bligaard, J. K. Nørskov, S. Dahl, J. Matthiesen, C. H. Christensen, and J. Sehested, J. Catal. 224, 206 (2004).