Graph Neural Networks (GNNs) are a popular approach for predicting graph structured data. As GNNs tightly entangle the input graph into the neural network structure, common explainable AI approaches are not applicable. To a large extent, GNNs have remained black-boxes for the user so far. In this paper, we show that GNNs can in fact be naturally explained using higher-order expansions, i.e. by identifying groups of edges that jointly contribute to the prediction. Practically, we find that such explanations can be extracted using a nested attribution scheme, where existing techniques such as layer-wise relevance propagation (LRP) can be applied at each step. The output is a collection of walks into the input graph that are relevant for the prediction. Our novel explanation method, which we denote by GNN-LRP, is applicable to a broad range of graph neural networks and lets us extract practically relevant insights on sentiment analysis of text data, structure-property relationships in quantum chemistry, and image classification.Index Terms-graph neural networks, higher-order explanations, layer-wise relevance propagation, explainable machine learning. ! INTRODUCTIONMany interesting structures found in scientific and industrial applications can be expressed as graphs. Examples are lattices in fluid modeling, molecular geometry, biological interaction networks, or social / historical networks. Graph neural networks (GNNs) [1], [2] have been proposed as a method to learn from observations in general graph structures and have found use in an ever growing number of applications [3]-[8]. While GNNs make useful predictions, they typically act as black-boxes, and it has neither been directly possible (1) to extract novel insight from the learned model nor (2) to verify that the model has made the intended use of the graph structure, e.g. that it has avoided Clever Hans phenomena [9].Explainable AI (XAI) is an emerging research area that aims to extract interpretable insights from trained ML models [10], [11]. So far, research has focused, for example, on full black-box models [12], [13], self-explainable models [14], [15], or deep neural networks [16], where in all cases, the prediction can be attributed to the input features. For a GNN, however, the graph being received as input is deeply
Insight into the relation between morphology and transport properties of organic semiconductors can be gained using multiscale simulations. Since computing electronic properties, such as the intermolecular transfer integral, using quantum chemical (QC) methods requires a high computational cost, existing models assume several approximations. A machine learning (ML)-based multiscale approach is presented that allows to simulate charge transport in organic semiconductors considering the static disorder within disordered crystals. By mapping fingerprints of dimers to their respective transfer integral, a kernel ridge regression ML algorithm for the prediction of charge transfer integrals is trained and evaluated. Since QC calculations of the electronic structure must be performed only once, the use of ML reduces the computation time radically, while maintaining the prediction error small. Transfer integrals predicted by ML are utilized for the computation of charge carrier mobilities using off-lattice kinetic Monte Carlo (kMC) simulations. Benefiting from the rapid performance of ML, microscopic processes can be described accurately without the need for phenomenological approximations. The multiscale system is tested with the well-known molecular semiconductor pentacene. The presented methodology allows reproducing the experimentally observed anisotropy of the mobility and enables a fast estimation of the impact of disorder.However, to this day, the main issue concerning organic semiconductors is the low charge carrier mobility compared to their inorganic counterpart, [5] which limits the operational speed and performance of electronic devices. Largest measured mobilities are in the range of 10 cm 2 V −1 s −1 for highly crystalline pentacene [6] and rubrene. [7] Due to the lack of insight into the structureproperties relationships, the design of new materials often relies on chemical intuition. This makes it difficult to identify promising materials with enhanced mobility. Thus, theoretical and numerical models are considered as promising pathways to increase the understanding of the relation between charge transport properties and structural morphologies within organic materials at the nanoscale. [8] Various techniques ranging from analytic approaches [9][10][11][12][13][14] to simulation methods such as molecular dynamics (MD) [15][16][17] and kinetic Monte Carlo (kMC) [18][19][20][21][22][23] are utilized to model the impact of molecular structures on transport properties such as the charge carrier mobility. The above approaches consider different length scales: analytic models provide an empirical picture of charge transport in disordered organic materials at the continuum scale, but they do not account for the different molecular structures; MD simulations yield insight into microscopic properties on an atomistic scale (≤ 1 nm), however it is not feasible to obtain mesoscopic transport properties due to the high computational cost; kMC allows to bridge different length scales by implicitly linking structural an...
SchNetPack is a versatile neural networks toolbox that addresses both the requirements of method development and application of atomistic machine learning. Version 2.0 comes with an improved data pipeline, modules for equivariant neural networks as well as a PyTorch implementation of molecular dynamics. An optional integration with PyTorch Lightning and the Hydra configuration framework powers a flexible command-line interface. This makes SchNetPack 2.0 easily extendable with custom code and ready for complex training task such as generation of 3d molecular structures.
In addition to the impressive predictive power of machine learning (ML) models, more recently, explanation methods have emerged that enable an interpretation of complex nonlinear learning models such as deep neural networks. Gaining a better understanding is especially important e.g. for safetycritical ML applications or medical diagnostics etc. While such Explainable AI (XAI) techniques have reached significant popularity for classifiers, so far little attention has been devoted to XAI for regression models (XAIR). In this review, we clarify the fundamental conceptual differences of XAI for regression and classification tasks, establish novel theoretical insights and analysis for XAIR, provide demonstrations of XAIR on genuine practical regression problems, and finally discuss the challenges remaining for the field.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.