Machine learning systems that rely on training data collected from untrusted sources are vulnerable to poisoning attacks, in which adversaries controlling some of the collected data are able to induce a corrupted model. In this paper, we consider poisoning attacks where there is an adversary who has a particular target classifier in mind and hopes to induce a classifier close to that target by adding as few poisoning points as possible. We propose an efficient poisoning attack based on online convex optimization. Unlike previous model-targeted poisoning attacks, our attack comes with provable convergence to any achievable target classifier. The distance from the induced classifier to the target classifier is inversely proportional to the square root of the number of poisoning points. We also provide a certified lower bound on the minimum number of poisoning points needed to achieve a given target classifier. We report on experiments showing our attack has performance that is similar to or better than the state-of-the-art attacks in terms of attack success rate and distance to the target model, while providing the advantages of provable convergence, and the efficiency benefits associated with being an online attack that can determine near-optimal poisoning points incrementally.

The belief revision literature has largely focussed on the issue of how to revise one's beliefs in the light of information regarding matters of fact. Here we turn to an important but comparatively neglected issue: How might one extend a revision operator to handle conditionals as input? Our approach to this question of 'conditional revision' is distinctive insofar as it abstracts from the controversial details of how to revise by factual sentences. We introduce a 'plug and play' method for uniquely extending any iterated belief revision operator to the conditional case. The flexibility of our approach is achieved by having the result of a conditional revision by a Ramsey Test conditional ('arrow') determined by that of a plain revision by its corresponding material conditional ('hook'). It is shown to satisfy a number of new constraints that are of independent interest.

Commercial companies that collect user data on a large scale have been the main beneficiaries of this trend since the success of deep learning techniques is directly proportional to the amount of data available for training. Massive data collection required for deep learning presents obvious privacy issues. Users personal, highly sensitive data such as photos and voice recordings are kept indefinitely by the companies that collect it. Users can neither delete it nor restrict the purposes for which it is used. So, data privacy has been a very important concern for governments and companies these days. It gives rise to a very interesting challenge since on the one hand, we are pushing further and further for high-quality models and accessible data, but on the other hand, we need to keep data safe from both intentional and accidental leakage. The more personal the data is it is more restricted it means some of the most important social issues cannot be addressed using machine learning because researchers do not have access to proper training data. But by learning how to machine learning that protects privacy we can make a huge difference in solving many social issues like curing disease etc. Deep neural networks are susceptible to various inference attacks as they remember information about their training data. In this chapter, we introduce differential privacy, which ensures that different kinds of statistical analyses dont compromise privacy and federated learning, training a machine learning model on a data to which we do not have access to.

However, most of these are heuristics in nature (Twala et al., 2008), tailored towards some specific tree induction algorithm Decision trees are a popular family of models (Chen & Guestrin, 2016; Prokhorenkova et al., 2018), due to their attractive properties such as interpretability or make strong distributional assumptions about the data, and ability to handle heterogeneous such as the feature distribution factorizing completely (e.g., data. Concurrently, missing data is a prevalent mean, median imputation (Rubin, 1976)) or according to the occurrence that hinders performance of machine tree structure (Quinlan, 1993). As many works have compared learning models. As such, handling missing data the most prominent ones in empirical studies (Batista in decision trees is a well studied problem. In & Monard, 2003; Saar-Tsechansky & Provost, 2007), there this paper, we tackle this problem by taking a is no clear winner and ultimately, the adoption of a particular probabilistic approach. At deployment time, we strategy in practice boils down to its availability in the use tractable density estimators to compute the ML libraries employed. "expected prediction" of our models. At learning time, we fine-tune parameters of already learned In this work, we tackle handling missing data in trees at trees by minimizing their "expected prediction both learning and deployment time from a principled probabilistic loss" w.r.t.

Understanding the evolutionary dynamics of reinforcement learning under multi-agent settings has long remained an open problem. While previous works primarily focus on 2-player games, we consider population games, which model the strategic interactions of a large population comprising small and anonymous agents. This paper presents a formal relation between stochastic processes and the dynamics of independent learning agents who reason based on the reward signals. Using a master equation approach, we provide a novel unified framework for characterising population dynamics via a single partial differential equation (Theorem 1). Through a case study involving Cross learning agents, we illustrate that Theorem 1 allows us to identify qualitatively different evolutionary dynamics, to analyse steady states, and to gain insights into the expected behaviour of a population. In addition, we present extensive experimental results validating that Theorem 1 holds for a variety of learning methods and population games.

Symbolic techniques based on Satisfiability Modulo Theory (SMT) solvers have been proposed for analyzing and verifying neural network properties, but their usage has been fairly limited owing to their poor scalability with larger networks. In this work, we propose a technique for combining gradient-based methods with symbolic techniques to scale such analyses and demonstrate its application for model explanation. In particular, we apply this technique to identify minimal regions in an input that are most relevant for a neural network's prediction. Our approach uses gradient information (based on Integrated Gradients) to focus on a subset of neurons in the first layer, which allows our technique to scale to large networks. The corresponding SMT constraints encode the minimal input mask discovery problem such that after masking the input, the activations of the selected neurons are still above a threshold. After solving for the minimal masks, our approach scores the mask regions to generate a relative ordering of the features within the mask. This produces a saliency map which explains "where a model is looking" when making a prediction. We evaluate our technique on three datasets - MNIST, ImageNet, and Beer Reviews, and demonstrate both quantitatively and qualitatively that the regions generated by our approach are sparser and achieve higher saliency scores compared to the gradient-based methods alone.

Formalisms inspired by Quantum theory have been used in Cognitive Science for decades. Indeed, Quantum-Like (QL) approaches provide descriptive features that are inherently suitable for perception, cognition, and decision processing. A preliminary study on the feasibility of a QL robot perception model has been carried out for a robot with limited sensing capabilities. In this paper, we generalize such a model for multi-sensory inputs, creating a multidimensional world representation directly based on sensor readings. Given a 3-dimensional case study, we highlight how this model provides a compact and elegant representation, embodying features that are extremely useful for modeling uncertainty and decision. Moreover, the model enables to naturally define query operators to inspect any world state, which answers quantifies the robot's degree of belief on that state.

Knowledge Graphs (KG) are gaining increasing attention in both academia and industry. Despite their diverse benefits, recent research have identified social and cultural biases embedded in the representations learned from KGs. Such biases can have detrimental consequences on different population and minority groups as applications of KG begin to intersect and interact with social spheres. This paper aims at identifying and mitigating such biases in Knowledge Graph (KG) embeddings. As a first step, we explore popularity bias -- the relationship between node popularity and link prediction accuracy. In case of node2vec graph embeddings, we find that prediction accuracy of the embedding is negatively correlated with the degree of the node. However, in case of knowledge-graph embeddings (KGE), we observe an opposite trend. As a second step, we explore gender bias in KGE, and a careful examination of popular KGE algorithms suggest that sensitive attribute like the gender of a person can be predicted from the embedding. This implies that such biases in popular KGs is captured by the structural properties of the embedding. As a preliminary solution to debiasing KGs, we introduce a novel framework to filter out the sensitive attribute information from the KG embeddings, which we call FAN (Filtering Adversarial Network). We also suggest the applicability of FAN for debiasing other network embeddings which could be explored in future work.

Kendall transformation is a conversion of an ordered feature into a vector of pairwise order relations between individual values. This way, it preserves ranking of observations and represents it in a categorical form. Such transformation allows for generalisation of methods requiring strictly categorical input, especially in the limit of small number of observations, when discretisation becomes problematic. In particular, many approaches of information theory can be directly applied to Kendall-transformed continuous data without relying on differential entropy or any additional parameters. Moreover, by filtering information to this contained in ranking, Kendall transformation leads to a better robustness at a reasonable cost of dropping sophisticated interactions which are anyhow unlikely to be correctly estimated. In bivariate analysis, Kendall transformation can be related to popular non-parametric methods, showing the soundness of the approach. The paper also demonstrates its efficiency in multivariate problems, as well as provides an example analysis of a real-world data.

Human trust in automation plays an essential role in interactions between humans and automation. While a lack of trust can lead to a human's disuse of automation, over-trust can result in a human trusting a faulty autonomous system which could have negative consequences for the human. Therefore, human trust should be calibrated to optimize human-machine interactions with respect to context-specific performance objectives. In this article, we present a probabilistic framework to model and calibrate a human's trust and workload dynamics during his/her interaction with an intelligent decision-aid system. This calibration is achieved by varying the automation's transparency--the amount and utility of information provided to the human. The parameterization of the model is conducted using behavioral data collected through human-subject experiments, and three feedback control policies are experimentally validated and compared against a non-adaptive decision-aid system. The results show that human-automation team performance can be optimized when the transparency is dynamically updated based on the proposed control policy. This framework is a first step toward widespread design and implementation of real-time adaptive automation for use in human-machine interactions. Automation has become prevalent in the everyday lives of humans. However, despite significant technological advancements, human supervision and intervention are still necessary in almost all sectors of automation, ranging from manufacturing and transportation to disaster-management and healthcare [1]. Therefore, we expect that the future will be built around human-agent collectives [2] that will require efficient and successful interaction and coordination between humans and machines. It is well established that to achieve this coordination, human trust in automation plays a central role [3]-[5]. For example, the benefits of automation are lost when humans override automation due to a fundamental lack of trust [3], [5], and accidents may occur due to human mistrust in such systems [6]. Therefore, trust should be appropriately calibrated to avoid disuse or misuse of automation [4].