State-of-the-art deep-learning systems use decision rules that are challenging for humans to model. Explainable AI (XAI) attempts to improve human understanding but rarely accounts for how people typically reason about unfamiliar agents. We propose explicitly modeling the human explainee via Bayesian Teaching, which evaluates explanations by how much they shift explainees' inferences toward a desired goal. We assess Bayesian Teaching in a binary image classification task across a variety of contexts. Absent intervention, participants predict that the AI's classifications will match their own, but explanations generated by Bayesian Teaching improve their ability to predict the AI's judgements by moving them away from this prior belief. Bayesian Teaching further allows each case to be broken down into sub-examples (here saliency maps). These sub-examples complement whole examples by improving error detection for familiar categories, whereas whole examples help predict correct AI judgements of unfamiliar cases.

Functional Magnetic Resonance Imaging (fMRI) maps cerebral activation in response to stimuli but this activation is often difficult to detect, especially in low-signal contexts and single-subject studies. Accurate activation detection can be guided by the fact that very few voxels are, in reality, truly activated and that activated voxels are spatially localized, but it is challenging to incorporate both these facts. We provide a computationally feasible and methodologically sound model-based approach, implemented in the R package MixfMRI, that bounds the a priori expected proportion of activated voxels while also incorporating spatial context. Results on simulation experiments for different levels of activation detection difficulty are uniformly encouraging. The value of the methodology in low-signal and single-subject fMRI studies is illustrated on a sports imagination experiment. Concurrently, we also extend the potential use of fMRI as a clinical tool to, for example, detect awareness and improve treatment in individual patients in persistent vegetative state, such as traumatic brain injury survivors.

Face masks have long been used in many areas of everyday life to protect against the inhalation of hazardous fumes and particles. They also offer an effective solution in healthcare for bi-directional protection against air-borne diseases. Wearing and positioning the mask correctly is essential for its function. Convolutional neural networks (CNNs) offer an excellent solution for face recognition and classification of correct mask wearing and positioning. In the context of the ongoing COVID-19 pandemic, such algorithms can be used at entrances to corporate buildings, airports, shopping areas, and other indoor locations, to mitigate the spread of the virus. These application scenarios impose major challenges to the underlying compute platform. The inference hardware must be cheap, small and energy efficient, while providing sufficient memory and compute power to execute accurate CNNs at a reasonably low latency. To maintain data privacy of the public, all processing must remain on the edge-device, without any communication with cloud servers. To address these challenges, we present a low-power binary neural network classifier for correct facial-mask wear and positioning. The classification task is implemented on an embedded FPGA, performing high-throughput binary operations. Classification can take place at up to ~6400 frames-per-second, easily enabling multi-camera, speed-gate settings or statistics collection in crowd settings. When deployed on a single entrance or gate, the idle power consumption is reduced to 1.6W, improving the battery-life of the device. We achieve an accuracy of up to 98% for four wearing positions of the MaskedFace-Net dataset. To maintain equivalent classification accuracy for all face structures, skin-tones, hair types, and mask types, the algorithms are tested for their ability to generalize the relevant features over all subjects using the Grad-CAM approach.

Machine learning systems are often trained using data collected from historical decisions. If past decisions were biased, then automated systems that learn from historical data will also be biased. We propose a black-box approach to identify and remove biased training data. Machine learning models trained on such debiased data (a subset of the original training data) have low individual discrimination, often 0%. These models also have greater accuracy and lower statistical disparity than models trained on the full historical data. We evaluated our methodology in experiments using 6 real-world datasets. Our approach outperformed seven previous approaches in terms of individual discrimination and accuracy.

The labelling of food products in the EU is regulated by the Food Information of Customers (FIC). Companies are required to provide the corresponding information regarding nutrients and allergens among others. With the rise of e-commerce more and more food products are sold online. There are often errors in the online product descriptions regarding the FIC-relevant information due to low data quality in the vendors' product data base. In this paper we propose a hybrid approach of both rule-based and machine learning to verify nutrient declaration and allergen labelling according to FIC requirements. Special focus is given to the problem of false negatives in allergen prediction since this poses a significant health risk to customers. Results show that a neural net trained on a subset of the ingredients of a product is capable of predicting the allergens contained with a high reliability.

For model privacy, local model parameters in federated learning shall be obfuscated before sent to the remote aggregator. This technique is referred to as \emph{secure aggregation}. However, secure aggregation makes model poisoning attacks, e.g., to insert backdoors, more convenient given existing anomaly detection methods mostly require access to plaintext local models. This paper proposes SAFELearning which supports backdoor detection for secure aggregation. We achieve this through two new primitives - \emph{oblivious random grouping (ORG)} and \emph{partial parameter disclosure (PPD)}. ORG partitions participants into one-time random subgroups with group configurations oblivious to participants; PPD allows secure partial disclosure of aggregated subgroup models for anomaly detection without leaking individual model privacy. SAFELearning is able to significantly reduce backdoor model accuracy without jeopardizing the main task accuracy under common backdoor strategies. Extensive experiments show SAFELearning reduces backdoor accuracy from $100\%$ to $8.2\%$ for ResNet-18 over CIFAR-10 when $10\%$ participants are malicious.

We develop high performance multilingualAbstract Meaning Representation (AMR) sys-tems by projecting English AMR annotationsto other languages with weak supervision. Weachieve this goal by bootstrapping transformer-based multilingual word embeddings, in partic-ular those from cross-lingual RoBERTa (XLM-R large). We develop a novel technique forforeign-text-to-English AMR alignment, usingthe contextual word alignment between En-glish and foreign language tokens. This wordalignment is weakly supervised and relies onthe contextualized XLM-R word embeddings.We achieve a highly competitive performancethat surpasses the best published results forGerman, Italian, Spanish and Chinese.

An emerging amount of intelligent applications have been developed with the surge of Machine Learning (ML). Deep Neural Networks (DNNs) have demonstrated unprecedented performance across various fields such as medical diagnosis and autonomous driving. While DNNs are widely employed in security-sensitive fields, they are identified to be vulnerable to Neural Trojan (NT) attacks that are controlled and activated by the stealthy trigger. We call this vulnerable model adversarial artificial intelligence (AI). In this paper, we target to design a robust Trojan detection scheme that inspects whether a pre-trained AI model has been Trojaned before its deployment. Prior works are oblivious of the intrinsic property of trigger distribution and try to reconstruct the trigger pattern using simple heuristics, i.e., stimulating the given model to incorrect outputs. As a result, their detection time and effectiveness are limited. We leverage the observation that the pixel trigger typically features spatial dependency and propose TAD, the first trigger approximation based Trojan detection framework that enables fast and scalable search of the trigger in the input space. Furthermore, TAD can also detect Trojans embedded in the feature space where certain filter transformations are used to activate the Trojan. We perform extensive experiments to investigate the performance of the TAD across various datasets and ML models. Empirical results show that TAD achieves a ROC-AUC score of 0:91 on the public TrojAI dataset 1 and the average detection time per model is 7:1 minutes.

The individual data collected throughout patient follow-up constitute crucial information for assessing the risk of a clinical event, and eventually for adapting a therapeutic strategy. Joint models and landmark models have been proposed to compute individual dynamic predictions from repeated measures to one or two markers. However, they hardly extend to the case where the complete patient history includes much more repeated markers possibly. Our objective was thus to propose a solution for the dynamic prediction of a health event that may exploit repeated measures of a possibly large number of markers. We combined a landmark approach extended to endogenous markers history with machine learning methods adapted to survival data. Each marker trajectory is modeled using the information collected up to landmark time, and summary variables that best capture the individual trajectories are derived. These summaries and additional covariates are then included in different prediction methods. To handle a possibly large dimensional history, we rely on machine learning methods adapted to survival data, namely regularized regressions and random survival forests, to predict the event from the landmark time, and we show how they can be combined into a superlearner. Then, the performances are evaluated by cross-validation using estimators of Brier Score and the area under the Receiver Operating Characteristic curve adapted to censored data. We demonstrate in a simulation study the benefits of machine learning survival methods over standard survival models, especially in the case of numerous and/or nonlinear relationships between the predictors and the event. We then applied the methodology in two prediction contexts: a clinical context with the prediction of death for patients with primary biliary cholangitis, and a public health context with the prediction of death in the general elderly population at different ages. Our methodology, implemented in R, enables the prediction of an event using the entire longitudinal patient history, even when the number of repeated markers is large. Although introduced with mixed models for the repeated markers and methods for a single right censored time-to-event, our method can be used with any other appropriate modeling technique for the markers and can be easily extended to competing risks setting.

In this work, we propose a robust Bayesian sparse learning algorithm based on Bayesian group Lasso with spike and slab priors for the discovery of partial differential equations with variable coefficients. Using the samples draw from the posterior distribution with a Gibbs sampler, we are able to estimate the values of coefficients, together with their standard errors and confidence intervals. Apart from constructing the error bars, uncertainty quantification can also be employed for designing new criteria of model selection and threshold setting. This enables our method more adjustable and robust in learning equations with time-dependent coefficients. Three criteria are introduced for model selection and threshold setting to identify the correct terms: the root mean square, total error bar, and group error bar. Moreover, three noise filters are integrated with the robust Bayesian sparse learning algorithm for better results with larger noise. Numerical results demonstrate that our method is more robust than sequential grouped threshold ridge regression and group Lasso in noisy situations through three examples.