Structured prediction requires searching over a combinatorial number of structures. To tackle it, we introduce SparseMAP, a new method for sparse structured inference, together with corresponding loss functions. SparseMAP inference is able to automatically select only a few global structures: it is situated between MAP inference, which picks a single structure, and marginal inference, which assigns probability mass to all structures, including implausible ones. Importantly, SparseMAP can be computed using only calls to a MAP oracle, hence it is applicable even to problems where marginal inference is intractable, such as linear assignment. Moreover, thanks to the solution sparsity, gradient backpropagation is efficient regardless of the structure. SparseMAP thus enables us to augment deep neural networks with generic and sparse structured hidden layers. Experiments in dependency parsing and natural language inference reveal competitive accuracy, improved interpretability, and the ability to capture natural language ambiguities, which is attractive for pipeline systems.

An interactive image retrieval system learns which images in the database belong to a user's query concept, by analyzing the example images and feedback provided by the user. The challenge is to retrieve the relevant images with minimal user interaction. In this work, we propose to solve this problem by posing it as a binary classification task of classifying all images in the database as being relevant or irrelevant to the user's query concept. Our method combines active learning with graph-based semi-supervised learning (GSSL) to tackle this problem. Active learning reduces the number of user interactions by querying the labels of the most informative points and GSSL allows to use abundant unlabeled data along with the limited labeled data provided by the user. To efficiently find the most informative point, we use an uncertainty sampling based method that queries the label of the point nearest to the decision boundary of the classifier. We estimate this decision boundary using our heuristic of adaptive threshold. To utilize huge volumes of unlabeled data we use an efficient approximation based method that reduces the complexity of GSSL from $O(n^3)$ to $O(n)$, making GSSL scalable. We make the classifier robust to the diversity and noisy labels associated with images in large databases by incorporating information from multiple modalities such as visual information extracted from deep learning based models and semantic information extracted from the WordNet. High F1 scores within few relevance feedback rounds in our experiments with concepts defined on AnimalWithAttributes and Imagenet (1.2 million images) datasets indicate the effectiveness and scalability of our approach.

Representation learning algorithms are designed to learn abstract features that characterize data. State representation learning (SRL) focuses on a particular kind of representation learning where learned features are in low dimension, evolve through time, and are influenced by actions of an agent. As the representation learned captures the variation in the environment generated by agents, this kind of representation is particularly suitable for robotics and control scenarios. In particular, the low dimension helps to overcome the curse of dimensionality, provides easier interpretation and utilization by humans and can help improve performance and speed in policy learning algorithms such as reinforcement learning. This survey aims at covering the state-of-the-art on state representation learning in the most recent years. It reviews different SRL methods that involve interaction with the environment, their implementations and their applications in robotics control tasks (simulated or real). In particular, it highlights how generic learning objectives are differently exploited in the reviewed algorithms. Finally, it discusses evaluation methods to assess the representation learned and summarizes current and future lines of research.

This indicates that even protected networks can be unexpectedly vulnerable. This is a crucial problem for this specific line of research because the primary concern of these studies are security threats. To tackle this crucial problem, we aim to develop defense methods with theoretical guarantees. Our goal is to ensure the lower bounds on the size of adversarial perturbations that networks can never be deceived for each input. We refer to these lower bounds as certified invariant radii, or simply, invariant radii. To make them available in broad applications, there are two fundamental requirements to their calculation methods: 1. the minimality of assumptions on network structures, 2. the computational tractability. However, many existing approaches require strong assumptions and massive computational costs. For example, we could not ensure perturbation invariance for some network structures such as wide residual networks [42], which have been commonly used in the evaluations of defense methods. This work tackled this problem and we provide a widely applicable, yet, highly scalable method to ensure large invariant radii. Our basic idea is to bound the size of adversarial perturbations that networks can never be deceived Even though the concept of using the Lipschitz constant has already appeared in Szegedy et al. [37], how much certifications they can provide has not been studied well. We show we can ensure significantly larger invariant radii compared to a recent computationally efficient counterpart [32]. However, the size of certified invariant radii can still be insufficient to be practically meaningful in some cases. We addressed this issue with a novel training procedure that further strengthen perturbation invariance.

Deep neural networks (DNNs) have been shown to be vulnerable to adversarial examples - malicious inputs which are crafted by the adversary to induce the trained model to produce erroneous outputs. This vulnerability has inspired a lot of research on how to secure neural networks against these kinds of attacks. Although existing techniques increase the robustness of the models against white-box attacks, they are ineffective against black-box attacks. To address the challenge of black-box adversarial attacks, we propose Adversarial Model Cascades (AMC); a framework that performs better than existing state-of-the-art defenses, in both black-box and white-box settings and is easy to integrate into existing set-ups. Our approach trains a cascade of models by injecting images crafted from an already defended proxy model, to improve the robustness of the target models against adversarial attacks. AMC provides an increase in robustness of 8.175% & 7.115% for white-box attacks and 30.218% & 4.717% for black-box, in comparison to defensive distillation and adversarial hardening. To the best of our knowledge, ours is the first work that aims to provide a defense mechanism that can improve robustness against multiple adversarial attacks simultaneously.

Gene expression data represents a unique challenge in predictive model building, because of the small number of samples $(n)$ compared to the huge amount of features $(p)$. This "$n<

Model-based methods for recommender systems have been studied to provide more precise results. In systems with large corpus, the amount of calculation for learnt model to predict all user-item pairs' preferences is tremendous, which makes the model difficult to be directly employed in recommendation candidate generation stage. To overcome the calculation barrier, models like matrix factorization can resort to inner product form (i.e., use the inner product of user and item's latent factors as the preference) and index like hashing to perform efficient approximate k-nearest neighbor search. However, other more expressive interaction forms between user and item features, e.g., interactions through advanced deep neural networks, are still prevented from large corpus recommendation because of the amount of calculation. In this paper, we focus on the problem how arbitrary advanced models can be introduced to generate recommendations from large corpus. We propose a novel tree-based method which can provide logarithmic complexity prediction w.r.t. corpus size with more expressive deep neural networks. The main idea of tree-based model is to predict user interests coarse-to-fine, by traversing tree nodes top-down and making decisions whether to pick up each node to user. Furthermore, we show that the tree structure can also be jointly learnt towards better compatible with user interests' distribution, to facilitate both training and prediction. Experiments in two large-scale real-world datasets indicate that the proposed model significantly outperforms traditional methods. And online A/B test results in Taobao display advertising platform prove the effectiveness of the tree-based deep model in production.

In real-world applications, it is conceivable that only positive and unlabeled (PU) data are available for training a classifier. For instance, in land-cover image classification, images of urban regions can be easily labeled, while images of non-urban regions are difficult to annotate due to high diversity of non-urban regions containing, e.g., forest, seas, grasses, and soil (Li et al., 2011). To cope with such situations, PU classification has been actively studied (Letouzey et al., 2000; Elkan and Noto, 2008; du Plessis et al., 2015), and the state-of-the-art method allows us to systematically train deep neural networks only from PU data (Kiryo et al., 2017). However, existing PU classification methods typically require an estimate of the class-prior probability, and their performance is sensitive to the quality of class-prior estimation (Kiryo et al., 2017). Although various class-prior estimation methods from PU data have been proposed so far (du Plessis and Sugiyama, 2014; Ramaswamy et al., 2016; Jain et al., 2016; du Plessis et al., 2017; Northcutt et al., 2017), accurate estimation of the class-prior is still highly challenging particularly for high-dimensional data.

Although aviation accidents are rare, safety incidents occur more frequently and require a careful analysis to detect and mitigate risks in a timely manner. Analyzing safety incidents using operational data and producing event-based explanations is invaluable to airline companies as well as to governing organizations such as the Federal Aviation Administration (FAA) in the United States. However, this task is challenging because of the complexity involved in mining multi-dimensional heterogeneous time series data, the lack of time-step-wise annotation of events in a flight, and the lack of scalable tools to perform analysis over a large number of events. In this work, we propose a precursor mining algorithm that identifies events in the multidimensional time series that are correlated with the safety incident. Precursors are valuable to systems health and safety monitoring and in explaining and forecasting safety incidents. Current methods suffer from poor scalability to high dimensional time series data and are inefficient in capturing temporal behavior. We propose an approach by combining multiple-instance learning (MIL) and deep recurrent neural networks (DRNN) to take advantage of MIL's ability to learn using weakly supervised data and DRNN's ability to model temporal behavior. We describe the algorithm, the data, the intuition behind taking a MIL approach, and a comparative analysis of the proposed algorithm with baseline models. We also discuss the application to a real-world aviation safety problem using data from a commercial airline company and discuss the model's abilities and shortcomings, with some final remarks about possible deployment directions.

Wiatowski and B\"olcskei, 2015, proved that deformation stability and vertical translation invariance of deep convolutional neural network-based feature extractors are guaranteed by the network structure per se rather than the specific convolution kernels and non-linearities. While the translation invariance result applies to square-integrable functions, the deformation stability bound holds for band-limited functions only. Many signals of practical relevance (such as natural images) exhibit, however, sharp and curved discontinuities and are, hence, not band-limited. The main contribution of this paper is a deformation stability result that takes these structural properties into account. Specifically, we establish deformation stability bounds for the class of cartoon functions introduced by Donoho, 2001.