But these scenarios depend upon an unanswered question: are machines intelligent to begin with? Computers are essentially logic machines that process digital information. But in a recent paper entitled "The Emperor of Strong AI Has No Clothes," physicist Robert K. Logan in Toronto and Adriana Braga in Rio de Janeiro argue that the dream of a super intelligence has limits that its adherents choose to ignore. The things the Singularity will never get right amount to a long list, to quote the two researchers: "… curiosity, imagination, intuition, emotions, passion, desires, pleasure, aesthetics, joy, purpose, objectives, goals, telos, values, morality, experience, wisdom, judgment, and even humor." A clever programmer can figure out how to get a computer to answer human questions like "How is your mother feeling?", "What does chocolate taste like?", and "Don't you just love fresh snow?"

Transformer architectures have become the model of choice in natural language processing and are now being introduced into computer vision tasks such as image classification, object detection, and semantic segmentation. However, in the field of human pose estimation, convolutional architectures still remain dominant. In this work, we present PoseFormer, a purely transformer-based approach for 3D human pose estimation in videos without convolutional architectures involved. Inspired by recent developments in vision transformers, we design a spatial-temporal transformer structure to comprehensively model the human joint relations within each frame as well as the temporal correlations across frames, then output an accurate 3D human pose of the center frame. We quantitatively and qualitatively evaluate our method on two popular and standard benchmark datasets: Human3.6M and MPI-INF-3DHP. Extensive experiments show that PoseFormer achieves state-of-the-art performance on both datasets. Code is available at \url{https://github.com/zczcwh/PoseFormer}

Label noise in multiclass classification is a major obstacle to the deployment of learning systems. However, unlike the widely used class-conditional noise (CCN) assumption that the noisy label is independent of the input feature given the true label, label noise in real-world datasets can be aleatory and heavily dependent on individual instances. In this work, we investigate the instance-dependent noise (IDN) model and propose an efficient approximation of IDN to capture the instance-specific label corruption. Concretely, noting the fact that most columns of the IDN transition matrix have only limited influence on the class-posterior estimation, we propose a variational approximation that uses a single-scalar confidence parameter. To cope with the situation where the mapping from the instance to its confidence value could vary significantly for two adjacent instances, we suggest using instance embedding that assigns a trainable parameter to each instance. The resulting instance-confidence embedding (ICE) method not only performs well under label noise but also can effectively detect ambiguous or mislabeled instances. We validate its utility on various image and text classification tasks.

In this work we address the problem of solving ill-posed inverse problems in imaging where the prior is a variational autoencoder (VAE). Specifically we consider the decoupled case where the prior is trained once and can be reused for many different log-concave degradation models without retraining. Whereas previous MAP-based approaches to this problem lead to highly non-convex optimization algorithms, our approach computes the joint (space-latent) MAP that naturally leads to alternate optimization algorithms and to the use of a stochastic encoder to accelerate computations. The resulting technique (JPMAP) performs Joint Posterior Maximization using an Autoencoding Prior. We show theoretical and experimental evidence that the proposed objective function is quite close to bi-convex. Indeed it satisfies a weak bi-convexity property which is sufficient to guarantee that our optimization scheme converges to a stationary point. We also highlight the importance of correctly training the VAE using a denoising criterion, in order to ensure that the encoder generalizes well to out-of-distribution images, without affecting the quality of the generative model. This simple modification is key to providing robustness to the whole procedure. Finally we show how our joint MAP methodology relates to more common MAP approaches, and we propose a continuation scheme that makes use of our JPMAP algorithm to provide more robust MAP estimates. Experimental results also show the higher quality of the solutions obtained by our JPMAP approach with respect to other non-convex MAP approaches which more often get stuck in spurious local optima.

Prediction tasks about students have practical significance for both student and college. Making multiple predictions about students is an important part of a smart campus. For instance, predicting whether a student will fail to graduate can alert the student affairs office to take predictive measures to help the student improve his/her academic performance. With the development of information technology in colleges, we can collect digital footprints which encode heterogeneous behaviors continuously. In this paper, we focus on modeling heterogeneous behaviors and making multiple predictions together, since some prediction tasks are related and learning the model for a specific task may have the data sparsity problem. To this end, we propose a variant of LSTM and a soft-attention mechanism. The proposed LSTM is able to learn the student profile-aware representation from heterogeneous behavior sequences. The proposed soft-attention mechanism can dynamically learn different importance degrees of different days for every student. In this way, heterogeneous behaviors can be well modeled. In order to model interactions among multiple prediction tasks, we propose a co-attention mechanism based unit. With the help of the stacked units, we can explicitly control the knowledge transfer among multiple tasks. We design three motivating behavior prediction tasks based on a real-world dataset collected from a college. Qualitative and quantitative experiments on the three prediction tasks have demonstrated the effectiveness of our model.

Identifying the research topics that best describe the scope of a scientific publication is a crucial task for editors, in particular because the quality of these annotations determine how effectively users are able to discover the right content in online libraries. For this reason, Springer Nature, the world's largest academic book publisher, has traditionally entrusted this task to their most expert editors. These editors manually analyse all new books, possibly including hundreds of chapters, and produce a list of the most relevant topics. Hence, this process has traditionally been very expensive, time-consuming, and confined to a few senior editors. For these reasons, back in 2016 we developed Smart Topic Miner (STM), an ontology-driven application that assists the Springer Nature editorial team in annotating the volumes of all books covering conference proceedings in Computer Science. Since then STM has been regularly used by editors in Germany, China, Brazil, India, and Japan, for a total of about 800 volumes per year. Over the past three years the initial prototype has iteratively evolved in response to feedback from the users and evolving requirements. In this paper we present the most recent version of the tool and describe the evolution of the system over the years, the key lessons learnt, and the impact on the Springer Nature workflow. In particular, our solution has drastically reduced the time needed to annotate proceedings and significantly improved their discoverability, resulting in 9.3 million additional downloads. We also present a user study involving 9 editors, which yielded excellent results in term of usability, and report an evaluation of the new topic classifier used by STM, which outperforms previous versions in recall and F-measure.

This paper focuses on the first. It encompasses knowledge representation and reasoning, with a focus here on (non-classical) reasoning (a second companion paper will focus on representation). The focus is on the act of reasoning that determines if some data can be seen (or interpreted) as belonging to a particular class, not on long chains of reasoning using diverse knowledge. A significant weakness of Artificial Intelligence (AI) systems relative to humans is the inability to apply existing knowledge to a new problem, or to a situation that varies from what they were programmed for or trained for (also called transfer ability in some contexts). This causes systems to fail to recognise objects or activities in new settings, or to fail to adapt skills to variations (Davis and Marcus, 2015; Ersen et al., 2017).

Multi-label classification (MC) is a standard machine learning problem in which a data point can be associated with a set of classes. A more challenging scenario is given by hierarchical multi-label classification (HMC) problems, in which every prediction must satisfy a given set of hard constraints expressing subclass relationships between classes. In this paper, we propose C-HMCNN(h), a novel approach for solving HMC problems, which, given a network h for the underlying MC problem, exploits the hierarchy information in order to produce predictions coherent with the constraints and to improve performance. Furthermore, we extend the logic used to express HMC constraints in order to be able to specify more complex relations among the classes and propose a new model CCN(h), which extends C-HMCNN(h) and is again able to satisfy and exploit the constraints to improve performance. We conduct an extensive experimental analysis showing the superior performance of both C-HMCNN(h) and CCN(h) when compared to state-of-the-art models in both the HMC and the general MC setting with hard logical constraints.

We propose a universal building block of Convolutional Neural Network (ConvNet) to improve the performance without any inference-time costs. The block is named Diverse Branch Block (DBB), which enhances the representational capacity of a single convolution by combining diverse branches of different scales and complexities to enrich the feature space, including sequences of convolutions, multi-scale convolutions, and average pooling. After training, a DBB can be equivalently converted into a single conv layer for deployment. Unlike the advancements of novel ConvNet architectures, DBB complicates the training-time microstructure while maintaining the macro architecture, so that it can be used as a drop-in replacement for regular conv layers of any architecture. In this way, the model can be trained to reach a higher level of performance and then transformed into the original inference-time structure for inference. DBB improves ConvNets on image classification (up to 1.9% higher top-1 accuracy on ImageNet), object detection and semantic segmentation. The PyTorch code and models are released at https://github.com/DingXiaoH/DiverseBranchBlock.

Learning continuously during all model lifetime is fundamental to deploy machine learning solutions robust to drifts in the data distribution. Advances in Continual Learning (CL) with recurrent neural networks could pave the way to a large number of applications where incoming data is non stationary, like natural language processing and robotics. However, the existing body of work on the topic is still fragmented, with approaches which are application-specific and whose assessment is based on heterogeneous learning protocols and datasets. In this paper, we organize the literature on CL for sequential data processing by providing a categorization of the contributions and a review of the benchmarks. We propose two new benchmarks for CL with sequential data based on existing datasets, whose characteristics resemble real-world applications. We also provide a broad empirical evaluation of CL and Recurrent Neural Networks in class-incremental scenario, by testing their ability to mitigate forgetting with a number of different strategies which are not specific to sequential data processing. Our results highlight the key role played by the sequence length and the importance of a clear specification of the CL scenario.