640 Hits in 8.0 sec

Using Attribution to Decode Dataset Bias in Neural Network Models for Chemistry [article]

Kevin McCloskey, Ankur Taly, Federico Monti, Michael P. Brenner, Lucy Colwell
2019 arXiv   pre-print
Here we show how the binding mechanism learned by deep neural network models can be interrogated, using a recently described attribution method.  ...  In light of our findings, we prescribe a test that checks for dataset bias given a hypothesis.  ...  Thoughts for practitioners. The recent machine learning revolution has led to great excitement regarding the use of neural networks in chemistry.  ... 
arXiv:1811.11310v3 fatcat:wcv5qn64vfb5xb6p7fi5nwkg4q

Using attribution to decode binding mechanism in neural network models for chemistry

Kevin McCloskey, Ankur Taly, Federico Monti, Michael P. Brenner, Lucy J. Colwell
2019 Proceedings of the National Academy of Sciences of the United States of America  
Here we show how the binding mechanism learned by deep neural network models can be interrogated, using a recently described attribution method.  ...  We first work with carefully constructed synthetic datasets, in which the molecular features responsible for "binding" are fully known.  ...  Thoughts for Practitioners. The recent machine learning revolution has led to great excitement regarding the use of neural networks in chemistry.  ... 
doi:10.1073/pnas.1820657116 pmid:31127041 pmcid:PMC6575176 fatcat:s5umtnupqjdo7jx2qlps7n3jtq

Orbital Graph Convolutional Neural Network for Material Property Prediction [article]

Mohammadreza Karamad, Rishikesh Magar, Yuting Shi, Samira Siahrostami, Ian D. Gates, Amir Barati Farimani
2020 arXiv   pre-print
Therefore, to develop robust machine learningmodels for material properties prediction, it is imperative to include features representing such chemical attributes.  ...  In addition, we embedded an encoder-decoder network into the OGCNN enabling it to learn important features among basic atomic (elemental features), orbital-orbital interactions, and topological features  ...  Acknowledgements The template for the preprint has been taken from: https://  ... 
arXiv:2008.06415v1 fatcat:bzjq54pmwfcidgt5r4y7ddutiq

Quantitative interpretation explains machine learning models for chemical reaction prediction and uncovers bias

Dávid Péter Kovács, William McCorkindale, Alpha A. Lee
2021 Nature Communications  
Additionally, we identify Clever Hans predictions where the correct prediction is reached for the wrong reason due to dataset bias.  ...  We develop a framework to attribute predicted reaction outcomes both to specific parts of reactants, and to reactions in the training set.  ...  Recently there were many methods developed and applied successfully for attributing the predictions of neural networks to parts of the input.  ... 
doi:10.1038/s41467-021-21895-w pmid:33727552 fatcat:l2aqqyvszfbhjhjhled5ffdms4

Generative chemistry: drug discovery with deep learning generative models [article]

Yuemin Bian, Xiang-Qun Xie
2020 arXiv   pre-print
The purpose of this paper is to review the latest advances in generative chemistry which relies on generative modeling to expedite the drug discovery process.  ...  Commonly used chemical databases, molecular representations, and tools in cheminformatics and machine learning are covered as the infrastructure for the generative chemistry.  ...  ACKNOWLEDGEMENTS Authors would like to acknowledge the funding support to the Xie laboratory from the NIH NIDA (P30 DA035778A1) and DOD (W81XWH-16-1-0490).  ... 
arXiv:2008.09000v1 fatcat:ivznoc4bsbfoderwr2ted76fiq

CAGE: Constrained deep Attributed Graph Embedding

Debora Nozza, Elisabetta Fersini, Enza Messina
2019 Information Sciences  
We validated our approach on two different benchmark datasets for node classification.  ...  In order to analyze these graphs, the primary challenge is to find an effective way to represent them by preserving both structural properties and node attribute information.  ...  Supplementary material Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.ins.2019.12.082 . CRediT authorship contribution statement  ... 
doi:10.1016/j.ins.2019.12.082 fatcat:3y7kp5wq7rfldmox6ir5w6y3mm

Categorical Normalizing Flows via Continuous Transformations [article]

Phillip Lippe, Efstratios Gavves
2021 arXiv   pre-print
Using a factorized decoder, we introduce an inductive bias to model any interactions in the normalizing flow.  ...  In this paper, we investigate Categorical Normalizing Flows, that is normalizing flows for categorical data.  ...  ACKNOWLEDGEMENTS We thank SURFsara for the support in using the Lisa Compute Cluster. REFERENCES John Adrian Bondy, Uppaluri Siva Ramachandra Murty, and others. 1976.  ... 
arXiv:2006.09790v3 fatcat:dwecz4ikjzhppirfzay57xy2cq

Assigning Confidence to Molecular Property Prediction [article]

AkshatKumar Nigam, Robert Pollice, Matthew F. D. Hurley, Riley J. Hickman, Matteo Aldeghi, Naruki Yoshikawa, Seyone Chithrananda, Vincent A. Voelz, Alán Aspuru-Guzik
2021 arXiv   pre-print
Computational modeling has rapidly advanced over the last decades, especially to predict molecular properties for chemistry, material science and drug design.  ...  First, our considerations for assessing confidence begin with dataset bias and size, data-driven property prediction and feature design.  ...  Like VAEs, they rely on a gaussian latent space to draw samples from and decode them using a neural network (referred to as generator network).  ... 
arXiv:2102.11439v1 fatcat:vblssmndlvchbjmubsnjjjppcy

Deep Generative Models for Materials Discovery and Machine Learning-Accelerated Innovation

Addis S. Fuhr, Bobby G. Sumpter
2022 Frontiers in Materials  
Machine learning and artificial intelligence (AI/ML) methods are beginning to have significant impact in chemistry and condensed matter physics.  ...  These approaches learn representations of a material structure and its corresponding chemistry or physics to accelerate materials discovery, which differs from traditional AI/ML methods that use statistical  ...  This work was carried out at Oak Ridge National Laboratory's Center for Nanophase Materials Sciences, a US Department of Energy Office of Science User Facility.  ... 
doi:10.3389/fmats.2022.865270 fatcat:fvp3ts6xcvaltpcfwywcz5bdg4

Deep Learning-Guided Surface Characterization for Autonomous Hydrogen Lithography [article]

Mohammad Rashidi, Jeremiah Croshaw, Kieran Mastel, Marcus Tamura, Hedieh Hosseinzadeh, Robert A. Wolkow
2019 arXiv   pre-print
We trained a convolutional neural network to locate and differentiate between surface features of the technologically relevant hydrogen-terminated silicon surface imaged using a scanning tunneling microscope  ...  Here we present an automation method for the identification of defects prior to atomic fabrication via hydrogen lithography using deep learning.  ...  S6 Confusion matrices showing labeling accuracy of the Test dataset for each of the 10 models.  ... 
arXiv:1902.08818v2 fatcat:4emwlxafz5gntfhvtpundqcjya

G2GT: Retrosynthesis Prediction with Graph to Graph Attention Neural Network and Self-Training [article]

Zaiyun Lin
2022 arXiv   pre-print
To solve this task, we propose a new graph-to-graph transformation model, G2GT, in which the graph encoder and graph decoder are built upon the standard transformer structure.  ...  Retrosynthesis prediction is one of the fundamental challenges in organic chemistry and related fields. The goal is to find reactants molecules that can synthesize product molecules.  ...  Graphormer Edge Encoding, Spatial Encoding, and Centrality Encoding are three Graphormer's main designs, which act as inductive biases in the neural network learning the graph representation.  ... 
arXiv:2204.08608v1 fatcat:e4b7pjtqezbi3nwdcfz7ljijku

Sequence-to-sequence modeling for graph representation learning

Aynaz Taheri, Kevin Gimpel, Tanya Berger-Wolf
2019 Applied Network Science  
Our methods use recurrent neural networks to encode and decode information from graph-structured data.  ...  Recurrent neural networks require sequences, so we choose several methods of traversing graphs using different types of substructures with various levels of granularity to generate sequences of nodes for  ...  Lee et al. (2018b) proposed Graph Attention Model (GAM) using recurrent neural networks for the graph classification task.  ... 
doi:10.1007/s41109-019-0174-8 fatcat:ftlwdepwrvachjj3hd3u3xz2eu

Chemulator: Fast, accurate thermochemistry for dynamical models through emulation

J. Holdship, S. Viti, T. J. Haworth, J. D. Ilee
2021 Astronomy and Astrophysics  
A neural network was then trained to map directly from inputs to outputs. Results.  ...  Chemical modelling serves two purposes in dynamical models: accounting for the effect of microphysics on the dynamics and providing observable signatures.  ...  The authors thank DiRAC for use of their HPC system which allowed this work to be performed. Article number, page 12 of 16 J.  ... 
doi:10.1051/0004-6361/202140357 fatcat:eunj7xa6ufexnex4ly5jxjk2na

Artificial Intelligence based Autonomous Molecular Design for Medical Therapeutic: A Perspective [article]

Rajendra P. Joshi, Neeraj Kumar
2021 arXiv   pre-print
Domain-aware machine learning (ML) models have been increasingly adopted for accelerating small molecule therapeutic design in the recent years.  ...  Our perspective serves as a guide for researchers to practice autonomous molecular design in therapeutic discovery.  ...  neural networks.  ... 
arXiv:2102.06045v1 fatcat:zbvqvi3hefh7tcapedwkbssopq

Exploration of Chemical Space by Machine Learning

Sergey Sosnin
Submitted to the Center for Computational and Data-Intensive Science and Engineering and Innovation on September 2020, in partial fulfillment of the requirements for the Doctoral Program in Computational  ...  Restricted Stochastic decoding In Subsection "Molecules decoding" we explained the basic decoding procedure. However, with neural networks, we use a restricted stochastic decoding technique.  ...  This network generates molecular structures biased towards the desired properties. A used-defined scoring function ( ) is used to control the generation.  ... 
doi:10.6084/m9.figshare.14160683.v1 fatcat:qqqvebwf3vbu5c2nte6a4uiowe
« Previous Showing results 1 — 15 out of 640 results