We propose a multi-armed bandit algorithm that explores based on randomizing its history. The key idea is to estimate the value of the arm from the bootstrap sample of its history, where we add pseudo observations after each pull of the arm. The pseudo observations seem to be harmful. But on the contrary, they guarantee that the bootstrap sample is optimistic with a high probability. Because of this, we call our algorithm Giro, which is an abbreviation for garbage in, reward out. We analyze Giro in a $K$-armed Bernoulli bandit and prove a $O(K \Delta^{-1} \log n)$ bound on its $n$-round regret, where $\Delta$ denotes the difference in the expected rewards of the optimal and best suboptimal arms. The main advantage of our exploration strategy is that it can be applied to any reward function generalization, such as neural networks. We evaluate Giro and its contextual variant on multiple synthetic and real-world problems, and observe that Giro is comparable to or better than state-of-the-art algorithms.

The paper proposes an on-line monitoring framework for continuous real-time safety/security in learning-based control systems (specifically application to a unmanned ground vehicle). We monitor validity of mappings from sensor inputs to actuator commands, controller-focused anomaly detection (CFAM), and from actuator commands to sensor inputs, system-focused anomaly detection (SFAM). CFAM is an image conditioned energy based generative adversarial network (EBGAN) in which the energy based discriminator distinguishes between proper and anomalous actuator commands. SFAM is based on an action condition video prediction framework to detect anomalies between predicted and observed temporal evolution of sensor data. We demonstrate the effectiveness of the approach on our autonomous ground vehicle for indoor environments and on Udacity dataset for outdoor environments.

ABSTRACT We consider technology-assisted mimicry attacks in the context of automatic speaker verification (ASV). We use ASV itself to select targeted speakers to be attacked by human-based mimicry. We recorded 6 naive mimics for whom we select target celebrities from VoxCeleb1 and VoxCeleb2 corpora (7,365 potential targets) using an i-vector system. The attacker attempts to mimic the selected target, with the utterances subjected to ASV tests using an independently developed x-vector system. Our main finding is negative: even if some of the attacker scores against the target speakers were slightly increased, our mimics did not succeed in spoofing the x-vector system. Interestingly, however, the relative ordering of the selected targets (closest, furthest, median) are consistent between the systems, which suggests some level of transferability between the systems.

Applying for life insurance is a long and often frustrating process. Thousands of questions on seemingly every medical condition ever suffered – except yours. "We've had multiple occurrences where people answer no to all the [medical] questions, then they come to the'other' box at the end and they'll go – 'oh yeah I've had X'. And that question is actually back there, but they didn't understand it so they defaulted to'other' and started writing chapter and verse about their medical condition," explains ANZ OnePath's chief underwriter Peter Tilocca. Whenever answers are given free form, typically the application will require the scrutiny of an underwriter.

We derive upper bounds on the generalization error of learning algorithms based on their \emph{algorithmic transport cost}: the expected Wasserstein distance between the output hypothesis and the output hypothesis conditioned on an input example. The bounds provide a novel approach to study the generalization of learning algorithms from an optimal transport view and impose less constraints on the loss function, such as sub-gaussian or bounded. We further provide several upper bounds on the algorithmic transport cost in terms of total variation distance, relative entropy (or KL-divergence), and VC dimension, thus further bridging optimal transport theory and information theory with statistical learning theory. Moreover, we also study different conditions for loss functions under which the generalization error of a learning algorithm can be upper bounded by different probability metrics between distributions relating to the output hypothesis and/or the input data. Finally, under our established framework, we analyze the generalization in deep learning and conclude that the generalization error in deep neural networks (DNNs) decreases exponentially to zero as the number of layers increases. Our analyses of generalization error in deep learning mainly exploit the hierarchical structure in DNNs and the contraction property of $f$-divergence, which may be of independent interest in analyzing other learning models with hierarchical structure.

A new app using artificial intelligence might be able to help detect autism in children earlier in their lives. A duo of lecturers at the Manukau Institute of Technology (MIT) have created the app that uses deep learning technology to identify autistic traits. Autism is a neurodevelopmental condition that affects cognitive, sensory, and social processing, changing the way people see the world and interact with others. Dr Fadi Fayez says the Autism AI (artificial intelligence) app asks 10 questions across four age groups - toddlers, children, adolescents and adults. READ MORE: * Artificial intelligence knows what you'll choose before you've made up your mind * Artificial intelligence is changing our lives and now is the time to decide how * This optical illusion could help diagnose people with autism * Women are being diagnosed with autism later in life An artificial neural network method processes the answers and shows whether to pursue a formal clinical diagnosis.

A global team of scientists have discovered AI can tell an individual's personality traits by analysing their eye movements. The scientists used machine learning to discover a link between eye movements and the personality of a person earlier this year. Armed with this information, the researchers deployed AI to analyse the eye movements of 42 students. The results were announced last week. In psychology, there are five basic personality traits: Extraversion, Agreeableness, Openness, Conscientiousness, and Neuroticism.

The increasing quantity of PV generation connected to distribution networks is creating challenges in maintaining and controlling voltages in those distribution networks. Determining the maximum hosting capacity for new PV installations based on the historical data is an essential task for distribution networks. Analyzing all historical data in large distribution networks is impractical. Therefore, this paper focuses on how to time efficiently identify the critical cases for evaluating the voltage impacts of the new large PV applications in medium voltage (MV) distribution networks. A systematic approach is proposed to cluster medium voltage nodes based on electrical adjacency and time blocks. MV nodes are clustered along with the voltage magnitudes and time blocks. Critical cases of each cluster can be used for further power flow study. This method is scalable and can time efficiently identify cases for evaluating PV investment on medium voltage networks.

Conditional kernel mean embeddings are nonparametric models that encode conditional expectations in a reproducing kernel Hilbert space. While they provide a flexible and powerful framework for probabilistic inference, their performance is highly dependent on the choice of kernel and regularization hyperparameters. Nevertheless, current hyperparameter tuning methods predominantly rely on expensive cross validation or heuristics that is not optimized for the inference task. For conditional kernel mean embeddings with categorical targets and arbitrary inputs, we propose a hyperparameter learning framework based on Rademacher complexity bounds to prevent overfitting by balancing data fit against model complexity. Our approach only requires batch updates, allowing scalable kernel hyperparameter tuning without invoking kernel approximations. Experiments demonstrate that our learning framework outperforms competing methods, and can be further extended to incorporate and learn deep neural network weights to improve generalization.

In less than 50 years, Artificial Intelligence (AI) will match humans on traits like adaptability, creativity and emotional intelligence, an expert has predicted. Speaking at the "Festival of Dangerous Ideas" at University of New South Wales in Sydney on Sunday, Professor Toby Walsh said AI will match human intelligence by 2062. "Toby Walsh, Scientia Professor of Artificial Intelligence at UNSW Sydney, has put a date on this looming reality. "He considers 2062 the year that artificial intelligence will match human intelligence, although a fundamental shift has already occurred in the world as we know it," the university said in a statement. Walsh argued that we are already experiencing the risks of AI that seem to be so far in the future.