In the past five years, the areas of adversarial attacks (Szegedy et al., 2013) on deep learning models, as well as defenses against such attacks, have received significant attention in the deep learning research community (Yuan et al., 2017; Akhtar & Mian, 2018). Defenses against adversarial attacks can be categorized into two main types. Approaches of the first type modify the net training procedures or architectures, usually in order to make the net compute a smooth function; see, for example (Shaham et al., 2015; Gu & Rigazio, 2014; Cisse et al., 2017; Papernot et al., 2016b). Defenses of the second type leave the training procedure and architecture unchanged, but rather modify the data, aiming to detect or remove adversarial perturbations often by smoothing the input data. For example, Guo et al. (2017) applied image transformations, such as total variance minimization and quilting to smooth input images.

Knowledge Transfer (KT) techniques tackle the problem of transferring the knowledge from a large and complex neural network into a smaller and faster one. However, existing KT methods are tailored towards classification tasks and they cannot be used efficiently for other representation learning tasks. In this paper a novel knowledge transfer technique, that is capable of training a student model that maintains the same amount of mutual information between the learned representation and a set of (possible unknown) labels as the teacher model, is proposed. Apart from outperforming existing KT techniques, the proposed method allows for overcoming several limitations of existing methods providing new insight into KT as well as novel KT applications, ranging from knowledge transfer from handcrafted feature extractors to {cross-modal} KT from the textual modality into the representation extracted from the visual modality of the data.

It is of high interest for a company to identify customers expected to bring the largest profit in the upcoming period. Knowing as much as possible about each customer is crucial for such predictions. However, their demographic data, preferences, and other information that might be useful for building loyalty programs is often missing. Additionally, modeling relations among different customers as a network can be beneficial for predictions at an individual level, as similar customers tend to have similar purchasing patterns. We address this problem by proposing a robust framework for structured regression on deficient data in evolving networks with a supervised representation learning based on neural features embedding. The new method is compared to several unstructured and structured alternatives for predicting customer behavior (e.g. purchasing frequency and customer ticket) on user networks generated from customer databases of two companies from different industries. The obtained results show $4\%$ to $130\%$ improvement in accuracy over alternatives when all customer information is known. Additionally, the robustness of our method is demonstrated when up to $80\%$ of demographic information was missing where it was up to several folds more accurate as compared to alternatives that are either ignoring cases with missing values or learn their feature representation in an unsupervised manner.

We show how well known rules of back propagation arise from a weighted combination of finite automata. By redefining a finite automata as a predictor we combine the set of all $k$-state finite automata using a weighted majority algorithm. This aggregated prediction algorithm can be simplified using symmetry, and we prove the equivalence of an algorithm that does this. We demonstrate that this algorithm is equivalent to a form of a back propagation acting in a completely connected $k$-node neural network. Thus the use of the weighted majority algorithm allows a bound on the general performance of deep learning approaches to prediction via known results from online statistics. The presented framework opens more detailed questions about network topology; it is a bridge to the well studied techniques of semigroup theory and applying these techniques to answer what specific network topologies are capable of predicting. This informs both the design of artificial networks and the exploration of neuroscience models.

Graphs are a fundamental abstraction for modeling relational data. However, graphs are discrete and combinatorial in nature, and learning representations suitable for machine learning tasks poses statistical and computational challenges. In this work, we propose Graphite an algorithmic framework for unsupervised learning of representations over nodes in a graph using deep latent variable generative models. Our model is based on variational autoencoders (VAE), and differs from existing VAE frameworks for data modalities such as images, speech, and text in the use of graph neural networks for parameterizing both the generative model (i.e., decoder) and inference model (i.e., encoder). The use of graph neural networks directly incorporates inductive biases due to the spatial, local structure of graphs directly in the generative model. Moreover, we draw novel connections between graph neural networks and approximate inference via kernel embeddings of distributions. We demonstrate empirically that Graphite outperforms state-of-the-art approaches for the tasks of density estimation, link prediction, and node classification on synthetic and benchmark datasets.

Animals execute goal-directed behaviours despite the limited range and scope of their sensors. To cope, they explore environments and store memories maintaining estimates of important information that is not presently available. Recently, progress has been made with artificial intelligence (AI) agents that learn to perform tasks from sensory input, even at a human level, by merging reinforcement learning (RL) algorithms with deep neural networks, and the excitement surrounding these results has led to the pursuit of related ideas as explanations of non-human animal learning. However, we demonstrate that contemporary RL algorithms struggle to solve simple tasks when enough information is concealed from the sensors of the agent, a property called "partial observability". An obvious requirement for handling partially observed tasks is access to extensive memory, but we show memory is not enough; it is critical that the right information be stored in the right format. We develop a model, the Memory, RL, and Inference Network (MERLIN), in which memory formation is guided by a process of predictive modeling. MERLIN facilitates the solution of tasks in 3D virtual reality environments for which partial observability is severe and memories must be maintained over long durations. Our model demonstrates a single learning agent architecture that can solve canonical behavioural tasks in psychology and neurobiology without strong simplifying assumptions about the dimensionality of sensory input or the duration of experiences.

Group equivariant and steerable convolutional neural networks (regular and steerable G-CNNs) have recently emerged as a very effective model class for learning from signal data such as 2D and 3D images, video, and other data where symmetries are present. In geometrical terms, regular G-CNNs represent data in terms of scalar fields ("feature channels"), whereas the steerable G-CNN can also use vector or tensor fields ("capsules") to represent data. In algebraic terms, the feature spaces in regular G-CNNs transform according to a regular representation of the group G, whereas the feature spaces in Steerable G-CNNs transform according to the more general induced representations of G. In order to make the network equivariant, each layer in a G-CNN is required to intertwine between the induced representations associated with its input and output space. In this paper we present a general mathematical framework for G-CNNs on homogeneous spaces like Euclidean space or the sphere. We show, using elementary methods, that the layers of an equivariant network are convolutional if and only if the input and output feature spaces transform according to an induced representation. This result, which follows from G.W. Mackey's abstract theory on induced representations, establishes G-CNNs as a universal class of equivariant network architectures, and generalizes the important recent work of Kondor & Trivedi on the intertwiners between regular representations.

Clustering and classification critically rely on distance metrics that provide meaningful comparisons between data points. We present mixed-integer optimization approaches to find optimal distance metrics that generalize the Mahalanobis metric extensively studied in the literature. Additionally, we generalize and improve upon leading methods by removing reliance on pre-designated "target neighbors," "triplets," and "similarity pairs." Another salient feature of our method is its ability to enable active learning by recommending precise regions to sample after an optimal metric is computed to improve classification performance. This targeted acquisition can significantly reduce computational burden by ensuring training data completeness, representativeness, and economy. We demonstrate classification and computational performance of the algorithms through several simple and intuitive examples, followed by results on real image and medical datasets.

The recently proposed self-ensembling methods have achieved promising results in deep semi-supervised learning, which penalize inconsistent predictions of unlabeled data under different perturbations. However, they only consider adding perturbations to each single data point, while ignoring the connections between data samples. In this paper, we propose a novel method, called Smooth Neighbors on Teacher Graphs (SNTG). In SNTG, a graph is constructed based on the predictions of the teacher model, i.e., the implicit self-ensemble of models. Then the graph serves as a similarity measure with respect to which the representations of "similar" neighboring points are learned to be smooth on the low-dimensional manifold. We achieve state-of-the-art results on semi-supervised learning benchmarks. The error rates are 9.89%, 3.99% for CIFAR-10 with 4000 labels, SVHN with 500 labels, respectively. In particular, the improvements are significant when the labels are fewer. For the non-augmented MNIST with only 20 labels, the error rate is reduced from previous 4.81% to 1.36%. Our method also shows robustness to noisy labels.

Model based iterative reconstruction (MBIR) algorithms for low-dose X-ray CT are computationally expensive. To address this problem, we recently proposed a deep convolutional neural network (CNN) for low-dose X-ray CT and won the second place in 2016 AAPM Low-Dose CT Grand Challenge. However, some of the texture were not fully recovered. To address this problem, here we propose a novel framelet-based denoising algorithm using wavelet residual network which synergistically combines the expressive power of deep learning and the performance guarantee from the framelet-based denoising algorithms. The new algorithms were inspired by the recent interpretation of the deep convolutional neural network (CNN) as a cascaded convolution framelet signal representation. Extensive experimental results confirm that the proposed networks have significantly improved performance and preserves the detail texture of the original images.