Accurate image segmentation plays a crucial role in medical image analysis, yet it faces great challenges of various shapes, diverse sizes, and blurry boundaries. To address these difficulties, square kernel-based encoder-decoder architecture has been proposed and widely used, but its performance remains still unsatisfactory. To further cope with these challenges, we present a novel double-branch encoder architecture. Our architecture is inspired by two observations: 1) Since the discrimination of features learned via square convolutional kernels needs to be further improved, we propose to utilize non-square vertical and horizontal convolutional kernels in the double-branch encoder, so features learned by the two branches can be expected to complement each other. 2) Considering that spatial attention can help models to better focus on the target region in a large-sized image, we develop an attention loss to further emphasize the segmentation on small-sized targets. Together, the above two schemes give rise to a novel double-branch encoder segmentation framework for medical image segmentation, namely Crosslink-Net. The experiments validate the effectiveness of our model on four datasets. The code is released at https://github.com/Qianyu1226/Crosslink-Net.

The policy gradient (PG) is one of the most popular methods for solving reinforcement learning (RL) problems. However, a solid theoretical understanding of even the "vanilla" PG has remained elusive for long time. In this paper, we apply recent tools developed for the analysis of SGD in non-convex optimization to obtain convergence guarantees for both REINFORCE and GPOMDP under smoothness assumption on the objective function and weak conditions on the second moment of the norm of the estimated gradient. When instantiated under common assumptions on the policy space, our general result immediately recovers existing $\widetilde{\mathcal{O}}(\epsilon^{-4})$ sample complexity guarantees, but for wider ranges of parameters (e.g., step size and batch size $m$) with respect to previous literature. Notably, our result includes the single trajectory case (i.e., $m=1$) and it provides a more accurate analysis of the dependency on problem-specific parameters by fixing previous results available in the literature. We believe that the integration of state-of-the-art tools from non-convex optimization may lead to identify a much broader range of problems where PG methods enjoy strong theoretical guarantees.

Machine learning models always make a prediction, even when it is likely to be inaccurate. This behavior should be avoided in many decision support applications, where mistakes can have severe consequences. Albeit already studied in 1970, machine learning with a reject option recently gained interest. This machine learning subfield enables machine learning models to abstain from making a prediction when likely to make a mistake. This survey aims to provide an overview on machine learning with a reject option. We introduce the conditions leading to two types of rejection, ambiguity and novelty rejection. Moreover, we define the existing architectures for models with a reject option, describe the standard learning strategies to train such models and relate traditional machine learning techniques to rejection. Additionally, we review strategies to evaluate a model's predictive and rejective quality. Finally, we provide examples of relevant application domains and show how machine learning with rejection relates to other machine learning research areas.

Deep learning models have gained great popularity in statistical modeling because they lead to very competitive regression models, often outperforming classical statistical models such as generalized linear models. The disadvantage of deep learning models is that their solutions are difficult to interpret and explain, and variable selection is not easily possible because deep learning models solve feature engineering and variable selection internally in a nontransparent way. Inspired by the appealing structure of generalized linear models, we propose a new network architecture that shares similar features as generalized linear models, but provides superior predictive power benefiting from the art of representation learning. This new architecture allows for variable selection of tabular data and for interpretation of the calibrated deep learning model, in fact, our approach provides an additive decomposition in the spirit of Shapley values and integrated gradients.

We combine beam search with the probabilistic pruning technique of nucleus sampling to create two deterministic nucleus search algorithms for natural language generation. The first algorithm, p-exact search, locally prunes the next-token distribution and performs an exact search over the remaining space. The second algorithm, dynamic beam search, shrinks and expands the beam size according to the entropy of the candidate's probability distribution. Despite the probabilistic intuition behind nucleus search, experiments on machine translation and summarization benchmarks show that both algorithms reach the same performance levels as standard beam search.

Over the last several years, end-to-end neural conversational agents have vastly improved in their ability to carry a chit-chat conversation with humans. However, these models are often trained on large datasets from the internet, and as a result, may learn undesirable behaviors from this data, such as toxic or otherwise harmful language. Researchers must thus wrestle with the issue of how and when to release these models. In this paper, we survey the problem landscape for safety for end-to-end conversational AI and discuss recent and related work. We highlight tensions between values, potential positive impact and potential harms, and provide a framework for making decisions about whether and how to release these models, following the tenets of value-sensitive design. We additionally provide a suite of tools to enable researchers to make better-informed decisions about training and releasing end-to-end conversational AI models.

A few curious cockatoos learned how to open residential waste bins in Australia, and now other birds have started copying them, with incidences of bin-looting spreading across eastern Australia in easily traceable waves. "If they had learned it individually, we would have seen this popping up randomly, but their method is really spreading from one suburb to the next," says Barbara Klump at the Max Planck Institute of Animal Behaviour in Germany. A few years ago, Richard Major at the Australian Museum Research Institute in Sydney filmed one of several sulphur-crested cockatoos (Cacatua galerita) lifting a bin lid, and he shared the video with Klump's colleague. Intrigued, the researchers asked suburbanites around Sydney and Wollongong to help them trace the phenomenon by reporting whether they saw, or didn't see, incidences of bin-looting in their neighbourhoods. When the team started the project in 2018, scientists had documented bin-opening by cockatoos in three suburbs.

We present Bayesian Controller Fusion (BCF): a hybrid control strategy that combines the strengths of traditional hand-crafted controllers and model-free deep reinforcement learning (RL). BCF thrives in the robotics domain, where reliable but suboptimal control priors exist for many tasks, but RL from scratch remains unsafe and data-inefficient. By fusing uncertainty-aware distributional outputs from each system, BCF arbitrates control between them, exploiting their respective strengths. We study BCF on two real-world robotics tasks involving navigation in a vast and long-horizon environment, and a complex reaching task that involves manipulability maximisation. For both these domains, there exist simple handcrafted controllers that can solve the task at hand in a risk-averse manner but do not necessarily exhibit the optimal solution given limitations in analytical modelling, controller miscalibration and task variation. As exploration is naturally guided by the prior in the early stages of training, BCF accelerates learning, while substantially improving beyond the performance of the control prior, as the policy gains more experience. More importantly, given the risk-aversity of the control prior, BCF ensures safe exploration and deployment, where the control prior naturally dominates the action distribution in states unknown to the policy. We additionally show BCF's applicability to the zero-shot sim-to-real setting and its ability to deal with out-of-distribution states in the real-world. BCF is a promising approach for combining the complementary strengths of deep RL and traditional robotic control, surpassing what either can achieve independently. The code and supplementary video material are made publicly available at https://krishanrana.github.io/bcf.

In order to construct an ethical artificial intelligence (AI) two complex problems must be overcome. Firstly, humans do not consistently agree on what is or is not ethical. Second, contemporary AI and machine learning methods tend to be blunt instruments which either search for solutions within the bounds of predefined rules, or mimic behaviour. An ethical AI must be capable of inferring unspoken rules, interpreting nuance and context, possess and be able to infer intent, and explain not just its actions but its intent. Using enactivism, semiotics, perceptual symbol systems and symbol emergence, we specify an agent that learns not just arbitrary relations between signs but their meaning in terms of the perceptual states of its sensorimotor system. Subsequently it can learn what is meant by a sentence and infer the intent of others in terms of its own experiences. It has malleable intent because the meaning of symbols changes as it learns, and its intent is represented symbolically as a goal. As such it may learn a concept of what is most likely to be considered ethical by the majority within a population of humans, which may then be used as a goal. The meaning of abstract symbols is expressed using perceptual symbols of raw sensorimotor stimuli as the weakest (consistent with Ockham's Razor) necessary and sufficient concept, an intensional definition learned from an ostensive definition, from which the extensional definition or category of all ethical decisions may be obtained. Because these abstract symbols are the same for both situation and response, the same symbol is used when either performing or observing an action. This is akin to mirror neurons in the human brain. Mirror symbols may allow the agent to empathise, because its own experiences are associated with the symbol, which is also associated with the observation of another agent experiencing something that symbol represents.

Healthcare representation learning on the Electronic Health Record (EHR) is seen as crucial for predictive analytics in the medical field. Many natural language processing techniques, such as word2vec, RNN and self-attention, have been adapted for use in hierarchical and time stamped EHR data, but fail when they lack either general or task-specific data. Hence, some recent works train healthcare representations by incorporating medical ontology (a.k.a. knowledge graph), by self-supervised tasks like diagnosis prediction, but (1) the small-scale, monotonous ontology is insufficient for robust learning, and (2) critical contexts or dependencies underlying patient journeys are never exploited to enhance ontology learning. To address this, we propose an end-to-end robust Transformer-based solution, Mutual Integration of patient journey and Medical Ontology (MIMO) for healthcare representation learning and predictive analytics. Specifically, it consists of task-specific representation learning and graph-embedding modules to learn both patient journey and medical ontology interactively. Consequently, this creates a mutual integration to benefit both healthcare representation learning and medical ontology embedding. Moreover, such integration is achieved by a joint training of both task-specific predictive and ontology-based disease typing tasks based on fused embeddings of the two modules. Experiments conducted on two real-world diagnosis prediction datasets show that, our healthcare representation model MIMO not only achieves better predictive results than previous state-of-the-art approaches regardless of sufficient or insufficient training data, but also derives more interpretable embeddings of diagnoses.