Quantum-chemical insights from deep tensor neural networks
By Kristof T. Schütt, Farhad Arbabzadah, Stefan Chmiela, K. Müller, Alexandre Tkatchenko. 2016
1,416 citations recorded by the source corpus.
Learning from data has led to paradigm shifts in a multitude of disciplines, including web, text and image search, speech recognition, as well as bioinformatics. Can machine learning enable similar breakthroughs in understanding quantum many-body systems? Here we develop an efficient deep learning approach that enables spatially and chemically resolved insights into quantum-mechanical observables of molecular systems. We unify concepts from many-body Hamiltonians with purpose-designed deep tensor neural networks, which leads to size-extensive and uniformly accurate (1 kcal mol−1) predictions in compositional and configurational chemical space for molecules of intermediate size. As an example of chemical relevance, the model reveals a classification of aromatic rings with respect to their stability. Further applications of our model for predicting atomic energies and local chemical potentials in molecules, reliable isomer energies, and molecules with peculiar electronic structure demonstrate the potential of machine learning for revealing insights into complex quantum-chemical systems. Machine learning is an increasingly popular approach to analyse data and make predictions. Here the authors develop a ‘deep learning’ framework for quantitative predictions and qualitative understanding of quantum-mechanical observables of chemical systems, beyond properties trivially contained in the training data.
Subfields: Computational Theory and Mathematics, Materials Chemistry, Molecular Biology, Computational Chemistry & Drug Discovery, AI for Physical Sciences
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Theory and Mathematics
- Risi Kondor
- Kyle Swanson
- Tommi S. Jaakkola
- Max Welling
- Jure Leskovec
- Charles Siegel
- Regina Barzilay
- Gisbert Schneider
In Materials Chemistry
- Tommi S. Jaakkola
- Regina Barzilay
- Risi Kondor
- Shitong Luo
- K. Müller
- Jian Tang
- Max Welling
- Frank No'e
In Molecular Biology
- Frank No'e
- Jure Leskovec
- Max Welling
- Jian Tang
- Tommi S. Jaakkola
- K. Müller
- Regina Barzilay
- Shitong Luo
In Computational Chemistry & Drug Discovery
- K. Müller
- Risi Kondor
- Frank No'e
- Grgoire Montavon
- Kyle Swanson
- Alán Aspuru-Guzik
- Alexandre Tkatchenko
- Wojciech Samek
In AI for Physical Sciences
- Risi Kondor
- Wojciech Samek
- K. Müller
- Kyle Swanson
- Frank No'e
- Shitong Luo
- Grgoire Montavon
- Brandon M. Anderson
Citations in the corpus
- E(n) Equivariant Graph Neural Networks (2021, methodology): “Schnet (Sch¨utt et al., 2017b) can be interpreted as an E( n ) invariant Graph Neural Network where φ cf receives as input relative distances and outputs a continuous filter convolution that multiplies the neighbor embeddings h .”
- A Survey on Explainable Artificial Intelligence (XAI): Toward Medical XAI (2019, methodology): “Deep Tensor Neural Network is used for quantum many-body systems [107].”
- Explainable Artificial Intelligence: Understanding, Visualizing and Interpreting Deep Learning Models (2017, background): “gists are rather interested in identifying the hidden laws of nature than just predicting some quantity with black box models. Thus, only models which are explainable are useful in this domain (c.f., [37, 32]). Compliance to legislation: AI systems are affecting more and more areas of our daily lif”
- Methods for interpreting and understanding deep neural networks (2017, methodology): “[58] proposed a deep tensor neural network model that incorporates sufficient structure and representational power to simultaneously achieve high predictive power and explainability.”
- Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation (2018, methodology): “In order to perform link prediction in G t ∪ C , our model first computes the node embeddings of an input graph using Graph Convolutional Networks (GCN) [20, 5, 18, 36, 8], a well-studied technique that achieves state-of-the-art performance in representation learning for molecules.”
- Learning Graph-Level Representation for Drug Discovery (2017, background): “(Yao et al. 2017; Schütt et al. 2017) apply neural network to predicate the molecular energy and reduced the predication error to 1 kcal/mol.”
- Covariant Compositional Networks For Learning Graphs (2018, background): “Other notable works on graph neural networks include (Li et al., 2015; Sch¨utt et al., 2017; Battaglia et al., 2016; Kipf & Welling, 2017).”
- SchNet: A continuous-filter convolutional neural network for modeling quantum interactions (2017, methodology): “The sequential updates from three interaction blocks allow SchNet to construct highly complex many-body representations in the spirit of DTNNs [20] while keeping rotational invariance due to the radial filters.”
- Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules (2016, background): “The Coulomb matrix approach, [21], scattering transforms [22], atomic distances [23], .”