Over the past decade, about 50 potential energy surfaces (PESs) for polyatomics with 4-11 atoms and clusters have been calculated using permutationally invariant polynomial method. This is a general, mainly linear least-squares method precise mathematical fitting of tens thousands electronic energies reactive nonreactive systems. A brief tutorial methodology given, including several recent improvements. Recent applications to formic acid dimer (the current record holder in size system),...

Conversational search is an emerging topic in the information retrieval community. One of major challenges to multi-turn conversational model conversation history answer current question. Existing methods either prepend turns question or use complicated attention mechanisms history. We propose a conceptually simple yet highly effective approach referred as embedding. It enables seamless integration into answering (ConvQA) built on BERT (Bidirectional Encoder Representations from...

Many model potential energy surfaces (PESs) have been reported for water; however, none are strictly from "first-principles". Here we report such a potential, based on many-body representation at the CCSD(T) level of theory up to four-body interaction. The new PES is benchmarked isomers water hexamer dissociation energies, harmonic frequencies, and unrestricted diffusion Monte Carlo (DMC) calculations zero-point energies Prism, Book, Cage isomers. Dissociation several 20-mer agree well with...

Conversational search is one of the ultimate goals information retrieval. Recent research approaches conversational by simplified settings response ranking and question answering, where an answer either selected from a given candidate set or extracted passage. These simplifications neglect fundamental role retrieval in search. To address this limitation, we introduce open-retrieval answering (ORConvQA) setting, learn to retrieve evidence large collection before extracting answers, as further...

Intelligent personal assistant systems with either text-based or voice-based conversational interfaces are becoming increasingly popular around the world. Retrieval-based conversation models have advantages of returning fluent and informative responses. Most existing studies in this area on open domain ''chit-chat'' conversations task / transaction oriented conversations. More research is needed for information-seeking There also a lack modeling external knowledge beyond dialog utterances...

Light-driven H2 generation using semiconductor nanocrystal heterostructures has attracted intense recent interest because of the ability to rationally improve their performance by tailoring size, composition, and morphology. In zero- one-dimensional nanomaterials, lifetime photoinduced charge-separated state is still too short for evolution reaction, limiting solar-to-H2 conversion efficiency. Here we report that two-dimensional (2D) CdS nanoplatelet (NPL)-Pt heterostructures, internal...

We report a full-dimensional, permutationally invariant potential energy surface (PES) for the cyclic formic acid dimer. This PES is least-squares fit to 13475 CCSD(T)-F12a/haTZ (VTZ H and aVTZ C O) energies. The energy-weighted, root-mean-square fitting error 11 cm-1 barrier double-proton transfer on 2848 cm-1, in good agreement with directly-calculated ab initio value of 2853 cm-1. zero-point vibrational 15 337 ± 7 obtained from diffusion Monte Carlo calculations. Energies fundamentals...

Since the discovery of roaming as an alternative molecular dissociation pathway in formaldehyde (H2CO), it has been indirectly observed numerous molecules. The phenomenon describes a frustrated with fragments at relatively large interatomic distances rather than following conventional transition-state dissociation; incipient radicals from parent molecule self-react to form products. Roaming identified spectroscopically through static product channel-resolved measurements, but not real-time...

The mathematical representation of large data sets electronic energies has seen substantial progress in the past 10 years. so-called Permutationally Invariant Polynomial (PIP) is one established approach. This approach dates from 2003, when a global potential energy surface (PES) for CH5+ was reported using basis polynomials that are invariant with respect to 120 permutations five equivalent H atoms. More recently, several approaches "machine learning" have been applied fit these sets....

Conversational question answering (ConvQA) is a simplified but concrete setting of conversational search. One its major challenges to leverage the conversation history understand and answer current question. In this work, we propose novel solution for ConvQA that involves three aspects. First, positional embedding method encode with position information using BERT in natural way. powerful technique text representation. Second, design attention mechanism (HAM) conduct "soft selection"...

Conversational assistants are being progressively adopted by the general population. However, they not capable of handling complicated information-seeking tasks that involve multiple turns information exchange. Due to limited communication bandwidth in conversational search, it is important for accurately detect and predict user intent conversations. In this paper, we investigate two aspects prediction an setting. First, extract features based on content, structural, sentiment...

Machine-learned potential energy surfaces (PESs) for molecules with more than 10 atoms are typically forced to use lower-level electronic structure methods such as density functional theory (DFT) and second-order Møller–Plesset perturbation (MP2). While these efficient realistic, they fall short of the accuracy "gold standard" coupled-cluster method, especially respect reaction isomerization barriers. We report a major step forward in applying Δ-machine learning method challenging case...

Permutationally invariant polynomial (PIP) regression has been used to obtain machine-learned potential energy surfaces, including analytical gradients, for many molecules and chemical reactions. Recently, the approach extended moderate size with up 15 atoms. The algorithm, "purification of basis," is computationally efficient energies; however, we found that recent extension despite being a remarkable advance over previous methods, could be further improved. Here, report developments...

We wish to describe a potential energy surface by using basis of permutationally invariant polynomials whose coefficients will be determined numerical regression so as smoothly fit dataset electronic energies well as, perhaps, gradients. The powers transformed internuclear distances, usually either Morse variables, exp(-ri,j/λ), where λ is constant range hyperparameter, or reciprocals the 1/ri,j. question we address how create most efficient basis, including (a) which keep discard, (b) many...

Assessments of machine-learning (ML) potentials are an important aspect the rapid development this field. We recently reported assessment linear-regression permutationally invariant polynomial (PIP) method for ethanol, using widely used (revised) rMD17 data set. demonstrated that PIP approach outperformed numerous other methods, e.g., ANI, PhysNet, sGDML, and p-KRR, with respect to precision notably speed [Houston et al., J. Chem. Phys. 2022, 156, 044120]. Here, we extend 21-atom aspirin...

A modified, computationally efficient method to provide permutationally invariant polynomial bases for molecular energy surface fitting via monomial symmetrization (Xie Z.; Bowman J. M. Chem. Theory Comput. 2010, 6, 26-34) is reported applications complex systems, characterized by many-body, non-covalent interactions. Two approaches, each able ensure the asymptotic zero-interaction limit of intrinsic potentials, are presented. They both based on tailored selection a subset polynomials...

Understanding and characterizing how people interact in information-seeking conversations is crucial developing conversational search systems. In this paper, we introduce a new dataset designed for purpose use it to analyze by user intent distribution, co-occurrence, flow patterns. The MSDialog labeled dialog of question answering (QA) interactions between information seekers providers from an online forum on Microsoft products. contains more than 2,000 multi-turn QA dialogs with 10,000...

Intelligent personal assistant systems that are able to have multi-turn conversations with human users becoming increasingly popular. Most previous research has been focused on using either retrieval-based or generation-based methods develop such systems. Retrieval-based the advantage of returning fluent and informative responses great diversity. However, performance is limited by size response repository. On other hand, can produce highly coherent any topics. But generated often generic not...

We describe software to incorporate electronic energies and gradients develop high-dimensional potential energy surfaces, using a permutationally invariant polynomial basis. The are obtained direct dynamics, the efficient B3LYP/6-31+G(d) level of theory. new is described along with extensive testing assessment benefits as well for CH4. Starting data set 9000 configurations, we examine training on sets only plus small 50. In addition standard root-mean-square fitting errors gradients,...

Tropolone, a 15-atom cyclic molecule, has received much interest both experimentally and theoretically due to its H-transfer tunneling dynamics. An accurate theoretical description is challenging owing the need develop high-level potential energy surface (PES) then simulate quantum-mechanical on this PES in full dimensionality. Here, we tackle aspects of challenge make detailed comparisons with experiments for numerous isotopomers. The PES, near CCSD(T)-quality, obtained using Δ-machine...

The potential energy surface of the methane-water dimer is represented as sum a new intrinsic two-body and pre-existing intramolecular potentials for monomers. Different fits CH4-H2O are reported. All these based on 30 467 ab initio interaction energies computed at CCSD(T)-F12b/haTZ (aug-cc-pVTZ C O, cc-pVTZ H) level theory. benchmark fit full-dimensional, permutationally-invariant analytical representation with root-mean-square (rms) fitting error 3.5 cm(-1). Two other computationally more...

We describe and apply a method to extend permutationally invariant polynomial (PIP) potential energy surface (PES) fitting molecules with more than 10 atoms. The creates compact basis of PIPs as the union obtained from fragments molecule. An application is reported for trans-N-methyl acetamide, where B3LYP/cc-pVDZ electronic energies gradients are used develop full-dimensional this prototype peptide performance several fragmented bases verified against benchmark PES using all (66) Morse...

An efficient method is described for generating a fragmented, permutationally invariant polynomial basis to fit electronic energies and, if available, gradients large molecules. The presented rests on the fragmentation of molecule into any number fragments while maintaining permutational invariance and uniqueness polynomials. new approach improves previous one reported by Qu Bowman avoiding repetition polynomials in fitting set speeding up gradient evaluations keeping accuracy PES....

``$Δ$-machine learning" refers to a machine learning approach bring property such as potential energy surface (PES) based on low-level (LL) density functional theory (DFT) energies and gradients close coupled cluster (CC) level of accuracy. Here we present an that uses the permutationally invariant polynomial (PIP) method fit high-dimensional PESs. The is represented by simple equation, in obvious notation $V_{LL{\rightarrow}CC}=V_{LL}+Δ{V_{CC-LL}}$, demonstrated for \ce{CH4}, \ce{H3O+},...