Showing papers on "Pairwise comparison published in 2020"

A review of fuzzy AHP methods for decision-making with subjective judgements

Abstract: Analytic Hierarchy Process (AHP) is a broadly applied multi-criteria decision-making method to determine the weights of criteria and priorities of alternatives in a structured manner based on pairwise comparison. As subjective judgments during comparison might be imprecise, fuzzy sets have been combined with AHP. This is referred to as fuzzy AHP or FAHP. An increasing amount of papers are published which describe different ways to derive the weights/priorities from a fuzzy comparison matrix, but seldomly set out the relative benefits of each approach so that the choice of the approach seems arbitrary. A review of various fuzzy AHP techniques is required to guide both academic and industrial experts to choose suitable techniques for a specific practical context. This paper reviews the literature published since 2008 where fuzzy AHP is applied to decision-making problems in industry, particularly the various selection problems. The techniques are categorised by the four aspects of developing a fuzzy AHP model: (i) representation of the relative importance for pairwise comparison, (ii) aggregation of fuzzy sets for group decisions and weights/priorities, (iii) defuzzification of a fuzzy set to a crisp value for final comparison, and (iv) consistency measurement of the judgements. These techniques are discussed in terms of their underlying principles, origins, strengths and weakness. Summary tables and specification charts are provided to guide the selection of suitable techniques. Tips for building a fuzzy AHP model are also included and six open questions are posed for future work.

A survey of blockchain consensus algorithms performance evaluation criteria

Abstract: How to reach an agreement in a blockchain network is a complex and important task that is defined as a consensus problem and has wide applications in reality including distributed computing, load balancing, and transaction validation in blockchains. Over recent years, many studies have been done to cope with this problem. In this paper, a comparative and analytical review on the state-of-the-art blockchain consensus algorithms is presented to enlighten the strengths and constraints of each algorithm. Based on their inherent specifications, each algorithm has a different domain of applicability that yields to propose several performance criteria for the evaluation of these algorithms. To overview and provide a basis of comparison for further work in the field, a set of incommensurable and conflicting performance evaluation criteria is identified and weighted by the pairwise comparison method. These criteria are classified into four categories including algorithms’ throughput, the profitability of mining, degree of decentralization and consensus algorithms vulnerabilities and security issues. Based on the proposed framework, the pros and cons of consensus algorithms are systematically analyzed and compared in order to provide a deep understanding of the existing research challenges and clarify the future study directions.

Poly-encoders: Architectures and Pre-training Strategies for Fast and Accurate Multi-sentence Scoring

Abstract: The use of deep pre-trained transformers has led to remarkable progress in a number of applications (Devlin et al., 2018). For tasks that make pairwise comparisons between sequences, matching a given input with a corresponding label, two approaches are common: Cross-encoders performing full self-attention over the pair and Bi-encoders encoding the pair separately. The former often performs better, but is too slow for practical use. In this work, we develop a new transformer architecture, the Poly-encoder, that learns global rather than token level self-attention features. We perform a detailed comparison of all three approaches, including what pre-training and fine-tuning strategies work best. We show our models achieve state-of-the-art results on four tasks; that Poly-encoders are faster than Cross-encoders and more accurate than Bi-encoders; and that the best results are obtained by pre-training on large datasets similar to the downstream tasks.

Consistency issues in the best worst method : Measurements and thresholds

Abstract: The Best-Worst Method (BWM) uses ratios of the relative importance of criteria in pairs based on the assessment done by decision-makers. When a decision-maker provides the pairwise comparisons in BWM, checking the acceptable inconsistency, to ensure the rationality of the assessments, is an important step. Although both the original and the extended versions of BWM have proposed several consistency measurements, there are some deficiencies, including: (i) the lack of a mechanism to provide immediate feedback to the decision-maker regarding the consistency of the pairwise comparisons being provided, (ii) the inability to consider the ordinal consistency into account, and (iii) the lack of consistency thresholds to determine the reliability of the results. To deal with these problems, this study starts by proposing a cardinal consistency measurement to provide immediate feedback, called the input-based consistency measurement, after which an ordinal consistency measurement is proposed to check the coherence of the order of the results (weights) against the order of the pairwise comparisons provided by the decision-maker. Finally, a method is proposed to balance cardinal consistency ratio under ordinal-consistent and ordinal-inconsistent conditions, to determine the thresholds for the proposed and the original consistency ratios.

Random walks on hypergraphs.

Abstract: In the past 20 years network science has proven its strength in modeling many real-world interacting systems as generic agents, the nodes, connected by pairwise edges. Nevertheless, in many relevant cases, interactions are not pairwise but involve larger sets of nodes at a time. These systems are thus better described in the framework of hypergraphs, whose hyperedges effectively account for multibody interactions. Here we propose and study a class of random walks defined on such higher-order structures and grounded on a microscopic physical model where multibody proximity is associated with highly probable exchanges among agents belonging to the same hyperedge. We provide an analytical characterization of the process, deriving a general solution for the stationary distribution of the walkers. The dynamics is ultimately driven by a generalized random-walk Laplace operator that reduces to the standard random-walk Laplacian when all the hyperedges have size 2 and are thus meant to describe pairwise couplings. We illustrate our results on synthetic models for which we have full control of the high-order structures and on real-world networks where higher-order interactions are at play. As the first application of the method, we compare the behavior of random walkers on hypergraphs to that of traditional random walkers on the corresponding projected networks, drawing interesting conclusions on node rankings in collaboration networks. As the second application, we show how information derived from the random walk on hypergraphs can be successfully used for classification tasks involving objects with several features, each one represented by a hyperedge. Taken together, our work contributes to unraveling the effect of higher-order interactions on diffusive processes in higher-order networks, shedding light on mechanisms at the heart of biased information spreading in complex networked systems.

Prioritizing the weights of the evaluation criteria under fuzziness: the fuzzy full consistency method – fucom-f

Abstract: Values, opinions, perceptions, and experiences are the forces that drive almost each and every kind of decision-making. Evaluation criteria are considered as sources of information used to compare alternatives and, as a result, make selection easier. Seeing their direct effect on the solution, weighting methods that most accurately determine criteria weights are needed. Unfortunately, the crisp values are insufficient to model real life problems due to the lack of complete information and the vagueness arising from linguistic assessments of decision-makers. Therefore, this paper proposes a novel subjective weighting method called the Fuzzy Full Consistency Method (FUCOM-F) for determining weights as accurately as possible under fuzziness. The most prominent feature of the proposed method is obtaining the most accurate weight values with very few pairwise comparisons. Consequently, thanks to this model, consistency and reliability of the results increase while the processing time and effort decrease. Moreover, an illustrative example related to the green supplier evaluation problem is performed. Finally, the robustness and effectiveness of the proposed fuzzy model is demonstrated by comparing it with fuzzy best-worst method (F-BWM) and fuzzy AHP (F-AHP) models.

Effective Diversity in Population Based Reinforcement Learning

Abstract: Exploration is a key problem in reinforcement learning, since agents can only learn from data they acquire in the environment. With that in mind, maintaining a population of agents is an attractive method, as it allows data be collected with a diverse set of behaviors. This behavioral diversity is often boosted via multi-objective loss functions. However, those approaches typically leverage mean field updates based on pairwise distances, which makes them susceptible to cycling behaviors and increased redundancy. In addition, explicitly boosting diversity often has a detrimental impact on optimizing already fruitful behaviors for rewards. As such, the reward-diversity trade off typically relies on heuristics. Finally, such methods require behavioral representations, often handcrafted and domain specific. In this paper, we introduce an approach to optimize all members of a population simultaneously. Rather than using pairwise distance, we measure the volume of the entire population in a behavioral manifold, defined by task-agnostic behavioral embeddings. In addition, our algorithm Diversity via Determinants (DvD), adapts the degree of diversity during training using online learning techniques. We introduce both evolutionary and gradient-based instantiations of DvD and show they effectively improve exploration without reducing performance when better exploration is not required.

Pairwise Fairness for Ranking and Regression.

Abstract: We present pairwise fairness metrics for ranking models and regression models that form analogues of statistical fairness notions such as equal opportunity, equal accuracy, and statistical parity. Our pairwise formulation supports both discrete protected groups, and continuous protected attributes. We show that the resulting training problems can be efficiently and effectively solved using existing constrained optimization and robust optimization techniques developed for fair classification. Experiments illustrate the broad applicability and trade-offs of these methods.

A Concentration Ratio for Nonlinear Best Worst Method

Abstract: Best Worst Method (BWM) is a multi-criteria decision-making method that is based on a structured pairwise comparison system. It uses two pairwise comparison vectors (best-to-others and others-to-wo...

Metrics for Graph Comparison: A Practitioner's Guide

Abstract: Comparison of graph structure is a ubiquitous task in data analysis and machine learning, with diverse applications in fields such as neuroscience, cyber security, social network analysis, and bioinformatics, among others. Discovery and comparison of structures such as modular communities, rich clubs, hubs, and trees yield insight into the generative mechanisms and functional properties of the graph. Often, two graphs are compared via a pairwise distance measure, with a small distance indicating structural similarity and vice versa. Common choices include spectral distances and distances based on node affinities. However, there has of yet been no comparative study of the efficacy of these distance measures in discerning between common graph topologies at different structural scales. In this work, we compare commonly used graph metrics and distance measures, and demonstrate their ability to discern between common topological features found in both random graph models and real world networks. We put forward a multi-scale picture of graph structure wherein we study the effect of global and local structures on changes in distance measures. We make recommendations on the applicability of different distance measures to the analysis of empirical graph data based on this multi-scale view. Finally, we introduce the Python library NetComp that implements the graph distances used in this work.

An empirical study on the perceived fairness of realistic, imperfect machine learning models

Abstract: There are many competing definitions of what statistical properties make a machine learning model fair. Unfortunately, research has shown that some key properties are mutually exclusive. Realistic models are thus necessarily imperfect, choosing one side of a trade-off or the other. To gauge perceptions of the fairness of such realistic, imperfect models, we conducted a between-subjects experiment with 502 Mechanical Turk workers. Each participant compared two models for deciding whether to grant bail to criminal defendants. The first model equalized one potentially desirable model property, with the other property varying across racial groups. The second model did the opposite. We tested pairwise trade-offs between the following four properties: accuracy; false positive rate; outcomes; and the consideration of race. We also varied which racial group the model disadvantaged. We observed a preference among participants for equalizing the false positive rate between groups over equalizing accuracy. Nonetheless, no preferences were overwhelming, and both sides of each trade-off we tested were strongly preferred by a non-trivial fraction of participants. We observed nuanced distinctions between participants considering a model "unbiased" and considering it "fair." Furthermore, even when a model within a trade-off pair was seen as fair and unbiased by a majority of participants, we did not observe consensus that a machine learning model was preferable to a human judge. Our results highlight challenges for building machine learning models that are perceived as fair and broadly acceptable in realistic situations.

Heterogeneous Graph Attention Network for Unsupervised Multiple-Target Domain Adaptation.

Abstract: Domain adaptation, which transfers the knowledge from label-rich source domain to unlabeled target domains, is a challenging task in machine learning. The prior domain adaptation methods focus on pairwise adaptation assumption with a single source and a single target domain, while little work concerns the scenario of one source domain and multiple target domains. Applying pairwise adaptation methods to this setting may be suboptimal, as they fail to consider the semantic association among multiple target domains. In this work we propose a deep semantic information propagation approach in the novel context of multiple unlabeled target domains and one labeled source domain. Our model aims to learn a unified subspace common for all domains with a heterogeneous graph attention network, where the transductive ability of the graph attention network can conduct semantic propagation of the related samples among multiple domains. In particular, the attention mechanism is applied to optimize the relationships of multiple domain samples for better semantic transfer. Then, the pseudo labels of the target domains predicted by the graph attention network are utilized to learn domain-invariant representations by aligning labeled source centroid and pseudo-labeled target centroid. We test our approach on four challenging public datasets, and it outperforms several popular domain adaptation methods.

A unifying mutual information view of metric learning: cross-entropy vs. pairwise losses

Abstract: Recently, substantial research efforts in Deep Metric Learning (DML) focused on designing complex pairwise-distance losses, which require convoluted schemes to ease optimization, such as sample mining or pair weighting. The standard cross-entropy loss for classification has been largely overlooked in DML. On the surface, the cross-entropy may seem unrelated and irrelevant to metric learning as it does not explicitly involve pairwise distances. However, we provide a theoretical analysis that links the cross-entropy to several well-known and recent pairwise losses. Our connections are drawn from two different perspectives: one based on an explicit optimization insight; the other on discriminative and generative views of the mutual information between the labels and the learned features. First, we explicitly demonstrate that the cross-entropy is an upper bound on a new pairwise loss, which has a structure similar to various pairwise losses: it minimizes intra-class distances while maximizing inter-class distances. As a result, minimizing the cross-entropy can be seen as an approximate bound-optimization (or Majorize-Minimize) algorithm for minimizing this pairwise loss. Second, we show that, more generally, minimizing the cross-entropy is actually equivalent to maximizing the mutual information, to which we connect several well-known pairwise losses. Furthermore, we show that various standard pairwise losses can be explicitly related to one another via bound relationships. Our findings indicate that the cross-entropy represents a proxy for maximizing the mutual information -- as pairwise losses do -- without the need for convoluted sample-mining heuristics. Our experiments over four standard DML benchmarks strongly support our findings. We obtain state-of-the-art results, outperforming recent and complex DML methods.

A Survey on Deep Hashing Methods

Abstract: Nearest neighbor search is to find the data points in the database such that the distances from them to the query are the smallest, which is a fundamental problem in various domains, such as computer vision, recommendation systems and machine learning. Hashing is one of the most widely used methods for its computational and storage efficiency. With the development of deep learning, deep hashing methods show more advantages than traditional methods. In this paper, we present a comprehensive survey of the deep hashing algorithms. Specifically, we categorize deep supervised hashing methods into pairwise similarity preserving, multiwise similarity preserving, implicit similarity preserving, classification-oriented preserving as well as quantization according to the manners of preserving the similarities. In addition, we also introduce some other topics such as deep unsupervised hashing and multi-modal deep hashing methods. Meanwhile, we also present some commonly used public datasets and the scheme to measure the performance of deep hashing algorithms. Finally, we discussed some potential research directions in conclusion.

A novel fuzzy hybrid neutrosophic decision‐making approach for the resilient supplier selection problem

Abstract: The objectives of this study are to mitigate the risk and disturbances to the supply chain, to offer required models for resolving the complex issues that arise, and to maintain the stability of the support system. Also, the uncertain conditions in a supply chain force decision-makers and experts to adopt a fuzzy-based evaluation platform to ensure secure and reliable consequences. The current study proposed a fuzzy neutrosophic decision-making approach for supplier evaluation and selection. The model is composed of a new weight aggregator that uses pairwise comparison, which has not been reported to date. The model uses a Dombi aggregator that is more qualified than other aggregators. The Dombi t-conorms and t-norms have the same properties as those of the general t-conorm and t-norm, which can enhance the flexibility of the information aggregation process via the adjustment of a parameter. A decision-making environment with uncertain condition and multiple factors is supposed. We applied this approach in a construction company to analyse the suppliers in a resilient supply chain management (RSCM) system using a MABAC (multi-attribute border approximation area comparison) tool. The accuracy of the proposed model was examined via sensitivity analysis tests. This study proposes a novel fuzzy-neutrosophic-based approach for resilient supplier selection. The main contributions of this research work are the design, implementation and analysis of a multi-attribute evaluation system with respect to fuzzy neutrosophic values. In this evaluation system, a new pairwise comparison is conducted with trapezoidal neutrosophic linguistic variables to determine the importance weights of supplier criteria. Typically, the provision of opinions regarding the qualitative performances of suppliers is a difficult and confusing responsibility for experts and supplier evaluators. Therefore, the propsed approach overcomes this problem by utilizing a pairwise comparison by neutrosophic values and proposes original Dombi aggregation operators for dealing with fuzzy neutrosophic sets.

Assessing Software Defection Prediction Performance: Why Using the Matthews Correlation Coefficient Matters

Abstract: Context: There is considerable diversity in the range and design of computational experiments to assess classifiers for software defect prediction. This is particularly so, regarding the choice of classifier performance metrics. Unfortunately some widely used metrics are known to be biased, in particular F1. Objective: We want to understand the extent to which the widespread use of the F1 renders empirical results in software defect prediction unreliable. Method: We searched for defect prediction studies that report both F1 and the Matthews correlation coefficient (MCC). This enabled us to determine the proportion of results that are consistent between both metrics and the proportion that change. Results: Our systematic review identifies 8 studies comprising 4017 pairwise results. Of these results, the direction of the comparison changes in 23% of the cases when the unbiased MCC metric is employed. Conclusion: We find compelling reasons why the choice of classification performance metric matters, specifically the biased and misleading F1 metric should be deprecated.

A Variational Bayesian Framework for Cluster Analysis in a Complex Network

Abstract: A complex network is a network with non-trivial topological structures. It contains not just topological information but also attribute information available in the rich content of nodes. Concerning the task of cluster analysis in a complex network, model-based algorithms are preferred over distance-based ones, as they avoid designing specific distance measures. However, their models are only applicable to complex networks where the attribute information is composed of attributes in binary form. To overcome this disadvantage, we introduce a three-layer node-attribute-value hierarchical structure to describe the attribute information in a flexible and interpretable manner. Then, a new Bayesian model is proposed to simulate the generative process of a complex network. In this model, the attribute information is generated by following the hierarchical structure while the links between pairwise nodes are generated by a stochastic blockmodel. To solve the corresponding inference problem, we develop a variational Bayesian algorithm called TARA, which allows us to identify functionally meaningful clusters through an iterative procedure. Our extensive experiment results show that TARA can be an effective algorithm for cluster analysis in a complex network. Moreover, the parallelized version of TARA makes it possible to perform efficiently at its tasks when applied to large complex networks.

Measuring performance of healthcare supply chains in India: A comparative analysis of multi-criteria decision making methods

Abstract: TThe supply chain forms the backbone of any organization. However, the effectiveness and efficiency of every activity get manifested in the financial outcome. Hence, measuring supply chain performance using financial metrics carries significance. The purpose of this paper is to carry out a comparative analysis of supply chain performances of leading healthcare organizations in India. In this regard, this paper presents an integrated multi-criteria decision making (MCDM) framework wherein we derive the weights of the criteria based on experts’ opinions using PIvot Pairwise RElative Criteria Importance Assessment (PIPRECIA) method. We then apply three distinct frameworks such as Multi-Attributive Border Approximation area Comparison (MABAC), Combined Compromise Solution (CoCoSo) and Measurement of alternatives and ranking according to COmpromise solution (MARCOS) for ranking purpose. In this context, this paper presents a comparative analysis of the results obtained from these approaches. The results show that large cap firms do not necessarily perform well. Further, the results of three MCDM frameworks demonstrates consistency.

Solving the supply chain problem using the best-worst method based on a novel Plithogenic model

Abstract: Plithogeny is the construction, development, and evolution of new entities from composition of contradictory (dissimilar) or non-contradictory multiple old entities. Using plithogenic aggregation operations may contribute to combining multiple opinions of decision-makers. The best-worst method (BWM) is an efficient multi-criteria decision-making (MCDM) technique that is based on pairwise comparison with the minimum number of comparisons, and on identification of the consistency of these comparisons. In this paper, we propose a plithogenic model based on the BWM. The result of plithogenic aggregation operation is taken as input to the BWM. We present two real-world supply chain problems as case studies to test the proposed model; these problems are warehouse location and plant evaluation. The results show the weight of each criterion, in order to arrange them based on their importance.

Base-criterion on multi-criteria decision-making method and its applications

Abstract: In this paper, Base-criterion method (BCM) is proposed to solve multi-criteria decision-making problems. According to BCM, instead of executing a pairwise comparisons between all criteria or execut...

A Differentiable Perceptual Audio Metric Learned from Just Noticeable Differences.

Abstract: Many audio processing tasks require perceptual assessment. The ``gold standard`` of obtaining human judgments is time-consuming, expensive, and cannot be used as an optimization criterion. On the other hand, automated metrics are efficient to compute but often correlate poorly with human judgment, particularly for audio differences at the threshold of human detection. In this work, we construct a metric by fitting a deep neural network to a new large dataset of crowdsourced human judgments. Subjects are prompted to answer a straightforward, objective question: are two recordings identical or not? These pairs are algorithmically generated under a variety of perturbations, including noise, reverb, and compression artifacts; the perturbation space is probed with the goal of efficiently identifying the just-noticeable difference (JND) level of the subject. We show that the resulting learned metric is well-calibrated with human judgments, outperforming baseline methods. Since it is a deep network, the metric is differentiable, making it suitable as a loss function for other tasks. Thus, simply replacing an existing loss (e.g., deep feature loss) with our metric yields significant improvement in a denoising network, as measured by subjective pairwise comparison.

Structural Patterns and Generative Models of Real-world Hypergraphs

Abstract: Graphs have been utilized as a powerful tool to model pairwise relationships between people or objects. Such structure is a special type of a broader concept referred to as hypergraph, in which each hyperedge may consist of an arbitrary number of nodes, rather than just two. A large number of real-world datasets are of this form - for example, lists of recipients of emails sent from an organization, users participating in a discussion thread or subject labels tagged in an online question. However, due to complex representations and lack of adequate tools, little attention has been paid to exploring the underlying patterns in these interactions. In this work, we empirically study a number of real-world hypergraph datasets across various domains. In order to enable thorough investigations, we introduce the multi-level decomposition method, which represents each hypergraph by a set of pairwise graphs. Each pairwise graph, which we refer to as a k-level decomposed graph, captures the interactions between pairs of subsets of k nodes. We empirically find that at each decomposition level, the investigated hypergraphs obey five structural properties. These properties serve as criteria for evaluating how realistic a hypergraph is, and establish a foundation for the hypergraph generation problem. We also propose a hypergraph generator that is remarkably simple but capable of fulfilling these evaluation metrics, which are hardly achieved by other baseline generator models.

GAMI-Net: An Explainable Neural Network based on Generalized Additive Models with Structured Interactions

Abstract: The lack of interpretability is an inevitable problem when using neural network models in real applications. In this paper, an explainable neural network based on generalized additive models with structured interactions (GAMI-Net) is proposed to pursue a good balance between prediction accuracy and model interpretability. GAMI-Net is a disentangled feedforward network with multiple additive subnetworks; each subnetwork consists of multiple hidden layers and is designed for capturing one main effect or one pairwise interaction. Three interpretability aspects are further considered, including a) sparsity, to select the most significant effects for parsimonious representations; b) heredity, a pairwise interaction could only be included when at least one of its parent main effects exists; and c) marginal clarity, to make main effects and pairwise interactions mutually distinguishable. An adaptive training algorithm is developed, where main effects are first trained and then pairwise interactions are fitted to the residuals. Numerical experiments on both synthetic functions and real-world datasets show that the proposed model enjoys superior interpretability and it maintains competitive prediction accuracy in comparison to the explainable boosting machine and other classic machine learning models.

Reconciling Qualitative, Abstract, and Scalable Modeling of Biological Networks

Abstract: Predicting biological systems' behaviors requires taking into account many molecular and genetic elements for which limited information is available past a global knowledge of their pairwise interactions. Logical modeling, notably with Boolean Networks (BNs), is a well-established approach that enables reasoning on the qualitative dynamics of networks. Several dynamical interpretations of BNs have been proposed. The synchronous and (fully) asynchronous ones are the most prominent, where the value of either all or only one component can change at each step. Here we prove that, besides being costly to analyze, these usual interpretations can preclude the prediction of certain behaviors observed in quantitative systems. We introduce an execution paradigm, the Most Permissive Boolean Networks (MPBNs), which offers the formal guarantee not to miss any behavior achievable by a quantitative model following the same logic. Moreover, MPBNs significantly reduce the complexity of dynamical analysis, enabling to model genome-scale networks.

How Much and When Do We Need Higher-order Information in Hypergraphs? A Case Study on Hyperedge Prediction

Abstract: Hypergraphs provide a natural way of representing group relations, whose complexity motivates an extensive array of prior work to adopt some form of abstraction and simplification of higher-order interactions. However, the following question has yet to be addressed: How much abstraction of group interactions is sufficient in solving a hypergraph task, and how different such results become across datasets? This question, if properly answered, provides a useful engineering guideline on how to trade off between complexity and accuracy of solving a downstream task. To this end, we propose a method of incrementally representing group interactions using a notion of n-projected graph whose accumulation contains information on up to n-way interactions, and quantify the accuracy of solving a task as n grows for various datasets. As a downstream task, we consider hyperedge prediction, an extension of link prediction, which is a canonical task for evaluating graph models. Through experiments on 15 real-world datasets, we draw the following messages: (a) Diminishing returns: small n is enough to achieve accuracy comparable with near-perfect approximations, (b) Troubleshooter: as the task becomes more challenging, larger n brings more benefit, and (c) Irreducibility: datasets whose pairwise interactions do not tell much about higher-order interactions lose much accuracy when reduced to pairwise abstractions.

PIGNet: A physics-informed deep learning model toward generalized drug-target interaction predictions.

Abstract: Recently, deep neural network (DNN)-based drug-target interaction (DTI) models are highlighted for their high accuracy with affordable computational costs. Yet, the models' insufficient generalization remains a challenging problem in the practice of in-silico drug discovery. We propose two key strategies to enhance generalization in the DTI model. The first one is to integrate physical models into DNN models. Our model, PIGNet, predicts the atom-atom pairwise interactions via physics-informed equations parameterized with neural networks and provides the total binding affinity of a protein-ligand complex as their sum. We further improved the model generalization by augmenting a wider range of binding poses and ligands to training data. PIGNet achieved a significant improvement in docking success rate, screening enhancement factor, and screening success rate by up to 2.01, 10.78, 14.0 times, respectively, compared to the previous DNN models. The physics-informed model also enables the interpretation of predicted binding affinities by visualizing the energy contribution of ligand substructures, providing insights for ligand optimization. Finally, we devised the uncertainty estimator of our model's prediction to qualify the outcomes and reduce the false positive rates.