Causality Learning Research Articles

An entropy-based causality framework for cross-level faults diagnosis and isolation in building HVAC systems

Faults, such as malfunctioning sensors, equipment, and control systems, significantly affect a building’s performance. Automatic fault detection and diagnosis (AFDD) tools have shown great potential in improving building performances, including both energy efficiency and indoor environment quality. Since modern buildings have integrated systems where multiple subsystems and equipment are coupled, many faults in a building are cross-level faults, i.e., faults occurring in one component that trigger operational abnormalities in other subsystems. Compared with non-cross-level faults, it is more challenging to isolate the root cause of a cross-level faults due to the system coupling effects. Bayesian networks (BNs) have been studied for the root cause isolation for building faults. While promising, existing BN-based diagnosis methods highly rely on expert domain knowledge, which is time-consuming and labor expensive, especially for cross-level faults. To address this challenge, we propose an entropy-based causality learning framework, termed Eigen-Entropy Causal Learning (EECL), to learn BN structures. The proposed method is data-driven without the use of expert domain knowledge; it utilizes causal inference to determine the causal mechanisms between faults status and symptoms to construct the BN model. To demonstrate the effectiveness of the proposed framework, three fault test cases are used for evaluation in this study. Experimental results show that the BN constructed by the proposed framework is able to conduct building cross-level faults diagnosis with a comparable isolation accuracy to those by domain knowledge while maintaining less complexed BN structure.

Federated Causality Learning with Explainable Adaptive Optimization

Discovering the causality from observational data is a crucial task in various scientific domains. With increasing awareness of privacy, data are not allowed to be exposed, and it is very hard to learn causal graphs from dispersed data, since these data may have different distributions. In this paper, we propose a federated causal discovery strategy (FedCausal) to learn the unified global causal graph from decentralized heterogeneous data. We design a global optimization formula to naturally aggregate the causal graphs from client data and constrain the acyclicity of the global graph without exposing local data. Unlike other federated causal learning algorithms, FedCausal unifies the local and global optimizations into a complete directed acyclic graph (DAG) learning process with a flexible optimization objective. We prove that this optimization objective has a high interpretability and can adaptively handle homogeneous and heterogeneous data. Experimental results on synthetic and real datasets show that FedCausal can effectively deal with non-independently and identically distributed (non-iid) data and has a superior performance.

Empowering brain cancer diagnosis: harnessing artificial intelligence for advanced imaging insights.

Artificial intelligence (AI) is increasingly being used in the medical field, specifically for brain cancer imaging. In this review, we explore how AI-powered medical imaging can impact the diagnosis, prognosis, and treatment of brain cancer. We discuss various AI techniques, including deep learning and causality learning, and their relevance. Additionally, we examine current applications that provide practical solutions for detecting, classifying, segmenting, and registering brain tumors. Although challenges such as data quality, availability, interpretability, transparency, and ethics persist, we emphasise the enormous potential of intelligent applications in standardising procedures and enhancing personalised treatment, leading to improved patient outcomes. Innovative AI solutions have the power to revolutionise neuro-oncology by enhancing the quality of routine clinical practice.

Causal Effect Estimation under Interference on Hypergraphs

Hypergraphs offer a powerful abstraction for representing multi-way group interactions, allowing hyperedges to connect any number of nodes. In contrast to prevailing approaches that focus on capturing statistical dependencies, our research explores hypergraphs from a causal perspective. Specifically, we tackle the problem of estimating individual treatment effects (ITE) on hypergraphs, aiming to determine the causal impact of interventions (e.g., wearing face covering) on outcomes (e.g., COVID-19 infection) for each individual node. Existing ITE estimation methods either assume no interference between individuals or consider interference only among connected individuals in regular graphs. However, such assumptions may not hold in real-world hypergraphs. Recognizing this, we propose a novel causality learning framework HyperSCI by modeling high-order interference on hyper-graphs. Through extensive experiments on real-world hypergraphs, we validate the effectiveness of HyperSCI and highlight the potential of causal inference in hypergraphs with complex group interactions. 1

Enabling causality learning in smart factories with hierarchical digital twins

Smart factories are complex systems where many different components need to interact and cooperate in order to achieve common goals. In particular, devices must be endowed with the skill of learning how to react in front of evolving situations and unexpected scenarios. In order to develop these capabilities, we argue that systems will need to build an internal, and possibly shared, representation of their operational world that represents causal relations between actions and observed variables. Within this context, digital twins will play a crucial role, by providing the ideal infrastructure for the standardisation and digitisation of the whole industrial process, laying the groundwork for the high-level learning and inference processes. In this paper, we introduce a novel hierarchical architecture enabled by digital twins, that can be exploited to build logical abstractions of the overall system, and to learn causal models of the environment directly from data. We implement our vision through a case study of a simulated production process. Our results in that scenario show that Bayesian networks and intervention via do-calculus can be effectively exploited within the proposed architecture to learn interpretable models of the environment. Moreover, we evaluate how the use of digital twins has a strong impact on the reduction of the physical complexity perceived by external applications.

CauRuler: Causal irredundant association rule miner for complex patient trajectory modelling

Background and Objectives.Discovering causal associations between variables is one of the main goals of clinical trials, with the ultimate aim of identifying the causes of specific health status. Prior knowledge of causal paths could help ensure patients do not develop the resultant conditions. In recent years, thanks to the enormous amount of health data stored with the support of digital tools, attempts have been made to employ Machine Learning to infer causality. Those methodologies suffer from some deficiencies in controlling cofounders when analysing causality, as well as providing causal rules general enough to be useful in healthcare practice. Conversely, this work presents and evaluates CauRuler, a new approach to deal with causality from association rules. The proposed approach uses a pruning strategy to reduce the association rule set, which does not compromise the causality learning capability of the algorithm. This behaviour makes the algorithm suitable for exploiting large health databases with thousands of patients and medical instances. CauRuler can control a larger number of confounders than other proposals, bringing robustness to causal analysis and avoiding the identification of spurious associations. Additionally, the method generalizes causality using anti-monotone properties to obtain complex and general causal paths. The method can target correct causal associations in complex medical databases with retrospective data. Method:CauRuler extends association rule mining with an irredundancy property so that the set of rules learnt is reduced in size and generalized. General association rules, conformed by fewer items, enable controlling more confounding variables to verify, with more statistical evidence on available data, if they represent causal paths in patient disease trajectories. Results:CauRuler has been tested on a complex real medical database (3,5 M visits to the primary care services between 2019 and 2020, and controlling over 15.000 different variables including diagnoses and demographic and other clinical patient data). The reduction of the rule set achieved by the pruning strategy goes from 7.732 to 2.240 rules, from which 46 have been found to have causality relationships in the patient trajectories, and generalized to 14 rules tested as true causal relationships thanks to the confounding analysis. These rules have been validated by clinicians with the support of a graphical map. The obtained causal paths control in average of 906 confounder variables, retrieving robust results. Conclusions:Causal relationships enable predicting causal paths between health conditions according to patient trajectories. Knowing these causal paths is crucial for understanding and preventing the appearance or worsening of diseases in patients. CauRuler, with high demanding thresholds, has proven its efficiency and effectiveness in targeting previously known causal associations between diagnoses, reaching consensus in the medical community. Softening these thresholds should help target interesting general causal paths.

Ensemble framework for causality learning with heterogeneous Directed Acyclic Graphs through the lens of optimization

Directed Acyclic Graphs (DAGs) are informative graphical outputs of causal learning algorithms to visualize the causal structure among variables. In practice, different causal learning algorithms are often used to establish a comprehensive analysis pool, which leads to the challenging problem of ensembling the heterogeneous DAGs with diverse and conflicting embedded information. While naive approaches such as MEDIAN and greedy heuristics are presented in the literature, a mathematical framework to build the ensemble DAG is not investigated. Therefore, we propose a two-stage ensemble framework for causality learning with heterogeneous DAGs. In the first stage, we implement a data partitioning procedure to categorize the input data. Then, we apply multiple causal learning algorithms to each class and ensemble the results across the partitions for each method. The results across different approaches in the second stage are then ensembled to produce the final DAG. We define a novel mathematical model based on the marginal contribution concept, reflecting the collective information for each edge from input graphs. We investigate the relationship between the proposed modeling approach and well-known problems such as rank aggregation and the Traveling Salesman Problem. We also present a parametric analysis for the distance function, enriched by insights from their asymptotic behavior. We design a simulation experiment to show the computational advantage of adopting a lazy constraints solution approach. In addition to the T-cell synthetic dataset, we use the dataset of water main breaks from 2017 to 2021 in the City of Tampa, Florida, to demonstrate the framework’s applicability. The results show that the optimal solution of the proposed framework outperforms current approaches in the literature based on causality-based Structural Intervention Distance performance measures.

Causal graph extraction from news: a comparative study of time-series causality learning techniques.

Causal graph extraction from news has the potential to aid in the understanding of complex scenarios. In particular, it can help explain and predict events, as well as conjecture about possible cause-effect connections. However, limited work has addressed the problem of large-scale extraction of causal graphs from news articles. This article presents a novel framework for extracting causal graphs from digital text media. The framework relies on topic-relevant variables representing terms and ongoing events that are selected from a domain under analysis by applying specially developed information retrieval and natural language processing methods. Events are represented as event-phrase embeddings, which make it possible to group similar events into semantically cohesive clusters. A time series of the selected variables is given as input to a causal structure learning techniques to learn a causal graph associated with the topic that is being examined. The complete framework is applied to the New York Times dataset, which covers news for a period of 246 months (roughly 20 years), and is illustrated through a case study. An initial evaluation based on synthetic data is carried out to gain insight into the most effective time-series causality learning techniques. This evaluation comprises a systematic analysis of nine state-of-the-art causal structure learning techniques and two novel ensemble methods derived from the most effective techniques. Subsequently, the complete framework based on the most promising causal structure learning technique is evaluated with domain experts in a real-world scenario through the use of the presented case study. The proposed analysis offers valuable insights into the problems of identifying topic-relevant variables from large volumes of news and learning causal graphs from time series.

Causality Learning with Wasserstein Generative Adversarial Networks

Conventional methods for causal structure learning from data face significant challenges due to combinatorial search space. Recently, the problem has been formulated into a continuous optimization framework with an acyclicity constraint to learn Directed Acyclic Graphs (DAGs). Such a framework allows the utilization of deep generative models for causal structure learning to better capture the relations between data sample distributions and DAGs. However, so far no study has experimented with the use of Wasserstein distance in the context of causal structure learning. Our model named DAG-WGAN combines the Wasserstein-based adversarial loss with an acyclicity constraint in an auto-encoder architecture. It simultaneously learns causal structures while improving its data generation capability. We compare the performance of DAG-WGAN with other models that do not involve the Wasserstein metric in order to identify its contribution to causal structure learning. Our model performs better with high cardinality data according to our experiments.

Time-Series Causality with Missing Data

Over the past years, researchers have proposed various methods to discover causal relationships among time-series data as well as algorithms to fill in missing entries in time-series data. Little to no work has been done in combining the two strategies for the purpose of learning causal relationships using unevenly sampled multivariate time-series data. In this paper, we examine how the causal parameters learnt from unevenly sampled data (with missing entries) deviates from the parameters learnt using the evenly sampled data (without missing entries). However, to obtain the causal relationship from a given time-series requires evenly sampled data, which suggests filling the missing data values before obtaining the causal parameters. Therefore, the proposed method is based on applying a Gaussian Process Regression (GPR) model for missing data recovery, followed by several pairwise Granger causality equations in Vector Autoregssive form to fit the recovered data and obtain the causal parameters. Experimental results show that the causal parameters generated by using GPR data filling offers much lower RMSE than the dummy model (fill with last seen entry) under all missing values percentage, suggesting that GPR data filling can better preserve the causal relationships when compared with dummy data filling, thus should be considered when dealing with unevenly sampled time-series causality learning.

Causality Learning from Time Series Data for the Industrial Finance Analysis via the Multi-dimensional Point Process

Causality Learning from Time Series Data for the Industrial Finance Analysis via the Multi-dimensional Point Process

Causality Learning from Time Series Data for the Industrial Finance Analysis via the Multi-dimensional Point Process

Causality Learning from Time Series Data for the Industrial Finance Analysis via the Multi-dimensional Point Process

Experience-based Causality Learning for Intelligent Agents

Understanding causality in text is crucial for intelligent agents. In this article, inspired by human causality learning, we propose an experience-based causality learning framework. Comparing to traditional approaches, which attempt to handle the causality problem relying on textual clues and linguistic resources, we are the first to use experience information for causality learning. Specifically, we first construct various scenarios for intelligent agents, thus, the agents can gain experience from interaction in these scenarios. Then, human participants build a number of training instances for agents of causality learning based on these scenarios. Each instance contains two sentences and a label. Each sentence describes an event that an agent experienced in a scenario, and the label indicates whether the sentence (event) pair has a causal relation. Accordingly, we propose a model that can infer the causality in text using experience by accessing the corresponding event information based on the input sentence pair. Experiment results show that our method can achieve impressive performance on the grounded causality corpus and significantly outperform the conventional approaches. Our work suggests that experience is very important for intelligent agents to understand causality.

Improving human understanding and design of complex multi-level systems with animation and parametric relationship supports

Complex systems are challenging to design, particularly when they contain multi-level organizations with non-obvious relationships among design components. Here, we investigate engineering students’ capacity to search for optimal nanoscale biosystem designs with stochastic component and system behaviors. The study aims to characterize information types that facilitate human learning and improve their complex system understanding and design proficiency. It is hypothesized that learning parametric system relationships and/or inter-level causal mechanisms improves design proficiency; these relationships and mechanisms are teachable through software interfaces. Two contrasting learning/design interfaces were developed that presented differing information types: an interface with performance charts that emphasized parametric relationship learning and an interface with agent-based animations that emphasized inter-level causality learning. Users improved on pre-/post-learning design tasks with both interfaces; users who demonstrated inter-level causal relationship understanding, which occurred primarily with the animation interface, had greater improvement. All users were then presented contrasting animations of systems with opposing emergent behaviors, resulting in many more participants demonstrating an understanding of inter-level causal behaviors. These findings reveal the difficulties in understanding and designing multi-level systems and that interactive software tools may convey crucial information that supports engineering design, particularly with respect to the development of reasoning skills for how system components relate across levels.

Software project risk analysis using Bayesian networks with causality constraints

Many risks are involved in software development and risk management has become one of the key activities in software development. Bayesian networks (BNs) have been explored as a tool for various risk management practices, including the risk management of software development projects. However, much of the present research on software risk analysis focuses on finding the correlation between risk factors and project outcome. Software project failures are often a result of insufficient and ineffective risk management. To obtain proper and effective risk control, risk planning should be performed based on risk causality which can provide more risk information for decision making. In this study, we propose a model using BNs with causality constraints (BNCC) for risk analysis of software development projects. Through unrestricted automatic causality learning from 302 collected software project data, we demonstrated that the proposed model can not only discover causalities in accordance with the expert knowledge but also perform better in prediction than other algorithms, such as logistic regression, C4.5, Naïve Bayes, and general BNs. This research presents the first causal discovery framework for risk causality analysis of software projects and develops a model using BNCC for application in software project risk management.

A Review of the Empirical Laws of Basic Learning in Pavlovian Conditioning

Contemporary learning research has provided multiple paradigms that have benefited not only researchers in the field, but also applied theorists and practitioners. However, the emphasis on theory development has made the learning literature almost impenetrable to nonexperts. In the present paper, we attempt to summarize not the different theoretical perspectives that have been proposed to explain different instances of learning, but the empirical relationships that testing of such theories has uncovered. Because the empirical relationships we summarize here hold across preparations and species, we suggest that such relationships should be understood as the empirical laws of basic learning. The focus of our review is the Pavlovian conditioning tradition, but most of these relationships also apply to instrumental learning and causality learning. We hope that the relatively novel organization we present here helps researchers and practitioners to directly incorporate these empirical principles into their current theoretical framework, whatever it may be.

Learning and Unlearning Mechanisms in Animats and Animals

This paper addresses the assumption, implicit in many robot and animat models of learning, that learning and unlearning are a pair of symmetrical processes. Unlearning mechanisms supposedly erase or remove existing learning because it is no longer relevant. Whether learning and unlearning result from the operation of symmetrical and antagonistic processes is an issue which has had a long but uneven history in animal and human psychology. This history is briefly recapitulated here. In particular, there is a contrast in the significance of the antagonistic processes model in the area of motivation compared to associative learning which has theoretical significance. For example, animat modelers frequently adopt a generic strengthening and weakening mechanism for all forms of learning and motivation representations without any consideration for its biological and psychological validity. In order to evaluate and question this, we examine the unlearning concept in a number of artefactual models drawn from a range of robotic and artificial life perspectives, and discuss their validity in terms of contemporary models of animal learning and motivation. Finally, we outline an alternative view of learning/unlearning, based on a recent contingency model of causality learning in humans, which does not rely on antagonistic processes and may have applications for artefacts.