Search

Deep learning models are receiving increasing attention in clinical decision-making, however the lack of interpretability and explainability impedes their deployment in day-to-day clinical practice. We propose REM, an interpretable and explainable methodology for extracting rules from deep neural networks and combining them with other data-driven and knowledge-driven rules. This allows integrating ma- chine learning and reasoning for investigating applied and basic biological research questions. We evaluate the utility of REM on the predictive tasks of classifying histological and immunohistochemical breast cancer subtypes from genotype and phenotype data. We demonstrate that REM efficiently extracts accurate, compre- hensible and, biologically relevant rulesets from deep neural networks that can be readily integrated with rulesets obtained from tree-based approaches. REM provides explanation facilities for predictions and enables the clinicians to vali- date and calibrate the extracted rulesets with their domain knowledge. With these functionalities, REM caters for a novel and direct human-in-the-loop approach in clinical decision making.

Multi-Agent Reinforcement Learning (MARL) encompasses a powerful class of methodologies that have been applied in a wide range of fields. An effective way to further empower these methodologies is to develop libraries and tools that could expand their interpretability and explainability. In this work, we introduce MARLeME: a MARL model extraction library, designed to improve explainability of MARL systems by approximating them with symbolic models. Symbolic models offer a high degree of interpretability, well-defined properties, and verifiable behaviour. Consequently, they can be used to inspect and better understand the underlying MARL system and corresponding MARL agents, as well as to replace all/some of the agents that are particularly safety and security critical., Presented at the KR2ML workshop at the 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada

International initiatives such as the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC), Cancer Genome Atlas (TCGA), and the International Cancer Genome Consortium (ICGC) are collecting multiple data sets at different genome-scales with the aim to identify novel cancer bio-markers and predict patient survival. To analyse such data, several machine learning, bioinformatics and statistical methods have been applied, among them neural networks such as autoencoders. Although these models provide a good statistical learning framework to analyse multi-omic and/or clinical data, there is a distinct lack of work on how to integrate diverse patient data and identify the optimal design best suited to the available data. In this paper, we investigate several autoencoder architectures that integrate a variety of cancer patient data types (e.g., multi-omics and clinical data). We perform extensive analyses of these approaches and provide a clear methodological and computational framework for designing systems that enable clinicians to investigate cancer traits and translate the results into clinical applications. We demonstrate how these networks can be designed, built and, in particular, applied to tasks of integrative analyses of heterogeneous breast cancer data. The results show that these approaches yield relevant data representations that, in turn, lead to accurate and stable diagnosis.

Autonomous software agents operating in dynamic environments need toconstantly reason about actions in pursuit of their goals, while taking intoconsideration norms which might be imposed on those actions. Normativepractical reasoning supports agents making decisions about what is best forthem to (not) do in a given situation. What makes practical reasoning chal-lenging is the interplay between goals that agents are pursuing and the normsthat the agents are trying to uphold. We offer a formalisation to allow agentsto plan for multiple goals and norms in the presence of durative actions thatcan be executed concurrently. We compare plans based on decision-theoreticnotions (i.e. utility) such that the utility gain of goals and utility loss of normviolations are the basis for this comparison. The set of optimal plans consistsof plans that maximise the overall utility, each of which can be chosen by theagent to execute. We provide an implementation of our proposal in AnswerSet Programming, thus allowing us to state the original problem in terms ofa logic program that can be queried for solutions with specific properties.

When formalising the rules of trust in the remote attesta- tion of TPM-based computer systems it is paramount that the rules are precisely understood, supporting unambiguous communication of infor- mation about system requirements between engineers. We present a dia- grammatic approach to modelling rules of trust using an extended version of concept diagrams. Within the context of our proof-of-concept Net- work Function Virtualisation and Attestation environment, these rules allow different level of trust to be explored and, importantly, allow us to identify when a computer system should not be trusted. To ensure that the modelling approach can be applied to general systems, we in- clude generic patterns for extending our domain model and rules of trust. Consequently, through the use of a formal, yet accessible, diagrammatic notation, domain experts can define rules of trust for their systems.

Research in psychology about reasoning has often been restricted to relatively inexpressive statements involving quantifiers (e.g. syllogisms). This is limited to situations that typically do not arise in practical settings, like ontology engineering. In order to provide an analysis of inference, we focus on reasoning tasks presented in external graphic representations where statements correspond to those involving multiple quantifiers and unary and binary relations. Our experiment measured participants’ performance when reasoning with two notations. The first notation used topological constraints to convey information via node-link diagrams (i.e. graphs). The second used topological and spatial constraints to convey information (Euler diagrams with additional graph-like syntax). We found that topo-spatial representations were more effective for inferences than topological representations alone. Reasoning with statements involving multiple quantifiers was harder than reasoning with single quantifiers in topological representations, but not in topo-spatial representations. These findings are compared to those in sentential reasoning tasks.

© Springer International Publishing AG 2017. Ontologies are notoriously hard to define, express and reason about. Many tools have been developed to ease the ontology debugging and reasoning, however they often lack accessibility and formalisation. A visual representation language, concept diagrams, was developed for expressing ontologies, which has been empirically proven to be cognitively more accessible to ontology users. In this paper we answer the question of “How can concept diagrams be used to reason about inconsistencies and incoherence of ontologies?". We do so by formalising a set of inference rules for concept diagrams that enables stepwise verification of the inconsistency and incoherence of a set of ontology axioms. The design of inference rules is driven by empirical evidence that concise (merged) diagrams are easier to comprehend for users than a set of lower level diagrams that are a one-to-one translation from OWL ontology axioms. We prove that our inference rules are sound, and exemplify how they can be used to reason about inconsistencies and incoherence.

High-tech systems are ubiquitous and often safety and security critical: reasoning about their correctness is paramount. Thus, precise modelling and formal reasoning are necessary in order to convey knowledge unambiguously and accurately. Whilst mathematical modelling adds great rigour, it is opaque to many stakeholders which leads to errors in data handling, delays in product release, for example. This is a major motivation for the development of diagrammatic approaches to formalisation and reasoning about models of knowledge. In this paper, we present an interactive theorem prover, called iCon, for a highly expressive diagrammatic logic that is capable of modelling OWL 2 ontologies and, thus, has practical relevance. Significantly, this work is the first to design diagrammatic inference rules using insights into what humans find accessible. Specifically, we conducted an experiment about relative cognitive benefits of primitive (small step) and derived (big step) inferences, and use the results to guide the implementation of inference rules in iCon.