Retrieving relevant targets from an extremely large target set under computational limits is a common challenge for information retrieval and recommendation systems. Tree models, which formulate targets as leaves of a tree with trainable node-wise scorers, have attracted a lot of interests in tackling this challenge due to their logarithmic computational complexity in both training and testing. Tree-based deep models (TDMs) and probabilistic label trees (PLTs) are two representative kinds of them. Though achieving many practical successes, existing tree models suffer from the training-testing discrepancy, where the retrieval performance deterioration caused by beam search in testing is not considered in training. This leads to an intrinsic gap between the most relevant targets and those retrieved by beam search with even the optimally trained node-wise scorers. We take a first step towards understanding and analyzing this problem theoretically, and develop the concept of Bayes optimality under beam search and calibration under beam search as general analyzing tools for this purpose. Moreover, to eliminate the discrepancy, we propose a novel algorithm for learning optimal tree models under beam search. Experiments on both synthetic and real data verify the rationality of our theoretical analysis and demonstrate the superiority of our algorithm compared to state-of-the-art methods.

Iterative machine teaching is a method for selecting an optimal teaching example that enables a student to efficiently learn a target concept at each iteration. Existing studies on iterative machine teaching are based on supervised machine learning and assume that there are teachers who know the true answers of all teaching examples. In this study, we consider an unsupervised case where such teachers do not exist; that is, we cannot access the true answer of any teaching example. Students are given a teaching example at each iteration, but there is no guarantee if the corresponding label is correct. Recent studies on crowdsourcing have developed methods for estimating the true answers from crowdsourcing responses. In this study, we apply these to iterative machine teaching for estimating the true labels of teaching examples along with student models that are used for teaching. Our method supports the collaborative learning of students without teachers. The experimental results show that the teaching performance of our method is particularly effective for low-level students in particular.

In building management, usually static thermal setpoints are used to maintain the inside temperature of a building at a comfortable level irrespective of its occupancy. This strategy can cause a massive amount of energy wastage and therewith increase energy related expenses. This paper explores how to optimise the setpoints used in a particular room during its unoccupied periods using machine learning approaches. We introduce a deep-learning model based on Recurrent Neural Networks (RNN) that can predict the temperatures of a future period directly where a particular room is unoccupied and by using these predicted temperatures, we define the optimal thermal setpoints to be used inside the room during the unoccupied period. We show that RNNs are particularly suitable for this learning task as they enable us to learn across many relatively short series, which is necessary to focus on particular operation modes of the air conditioning (AC) system. We evaluate the prediction accuracy of our RNN model against a set of state-of-the-art models and are able to outperform those by a large margin. We furthermore analyse the usage of our RNN model in optimising the energy consumption of an AC system in a real-world scenario using the temperature data from a university lecture theatre. Based on the simulations, we show that our RNN model can lead to savings around 20% compared with the traditional temperature controlling model that does not use optimisation techniques.

Prior work on continual learning often operate in a "task-aware" manner, by assuming that the task boundaries and identifies of the data instances are known at all times. While in practice, it is rarely the case that such information are exposed to the methods (i.e., thus called "task-free")--a setting that is relatively underexplored. Recent attempts on task-free continual learning build on previous memory replay methods and focus on developing memory management strategies such that model performance over priorly seen instances can be best retained. In this paper, looking from a complementary angle, we propose a principled approach to "edit" stored examples which aims to carry more updated information from the data stream in the memory. We use gradient updates to edit stored examples so that they are more likely to be forgotten in future updates. Experiments on five benchmark datasets show the proposed method can be seamlessly combined with baselines to significantly improve the performance. Code has been released at https://github.com/INK-USC/GMED.

Data augmentation have been intensively used in training deep neural network to improve the generalization, whether in original space (e.g., image space) or representation space. Although being successful, the connection between the synthesized data and the original data is largely ignored in training, without considering the distribution information that the synthesized samples are surrounding the original sample in training. Hence, the behavior of the network is not optimized for this. However, that behavior is crucially important for generalization, even in the adversarial setting, for the safety of the deep learning system. In this work, we propose a framework called Stochastic Batch Augmentation (SBA) to address these problems. SBA stochastically decides whether to augment at iterations controlled by the batch scheduler and in which a ''distilled'' dynamic soft label regularization is introduced by incorporating the similarity in the vicinity distribution respect to raw samples. The proposed regularization provides direct supervision by the KL-Divergence between the output soft-max distributions of original and virtual data. Our experiments on CIFAR-10, CIFAR-100, and ImageNet show that SBA can improve the generalization of the neural networks and speed up the convergence of network training.

Anchor-based techniques reduce the computational complexity of spectral clustering algorithms. Although empirical tests have shown promising results, there is currently a lack of theoretical support for the anchoring approach. We define a specific anchor-based algorithm and show that it is amenable to rigorous analysis, as well as being effective in practice. We establish the theoretical consistency of the method in an asymptotic setting where data is sampled from an underlying continuous probability distribution. In particular, we provide sharp asymptotic conditions for the algorithm parameters which ensure that the anchor-based method can recover with high probability disjoint clusters that are mutually separated by a positive distance. We illustrate the performance of the algorithm on synthetic data and explain how the theoretical convergence analysis can be used to inform the practical choice of parameter scalings. We also test the accuracy and efficiency of the algorithm on two large scale real data sets. We find that the algorithm offers clear advantages over standard spectral clustering. We also find that it is competitive with the state-of-the-art LSC method of Chen and Cai (Twenty-Fifth AAAI Conference on Artificial Intelligence, 2011), while having the added benefit of a consistency guarantee.

Global optimization of aerodynamic shapes usually requires a large number of expensive computational fluid dynamics simulations because of the high dimensionality of the design space. One approach to combat this problem is to reduce the design space dimension by obtaining a new representation. This requires a parametric function that compactly and sufficiently describes useful variation in shapes. We propose a deep generative model, B\'ezier-GAN, to parameterize aerodynamic designs by learning from shape variations in an existing database. The resulted new parameterization can accelerate design optimization convergence by improving the representation compactness while maintaining sufficient representation capacity. We use the airfoil design as an example to demonstrate the idea and analyze B\'ezier-GAN's representation capacity and compactness. Results show that B\'ezier-GAN both (1) learns smooth and realistic shape representations for a wide range of airfoils and (2) empirically accelerates optimization convergence by at least two times compared to state-of-the-art parameterization methods.

High-dimensional generative models have many applications including image compression, multimedia generation, anomaly detection and data completion. State-of-the-art estimators for natural images are autoregressive, decomposing the joint distribution over pixels into a product of conditionals parameterized by a deep neural network, e.g. a convolutional neural network such as the PixelCNN. However, PixelCNNs only model a single decomposition of the joint, and only a single generation order is efficient. For tasks such as image completion, these models are unable to use much of the observed context. To generate data in arbitrary orders, we introduce LMConv: a simple modification to the standard 2D convolution that allows arbitrary masks to be applied to the weights at each location in the image. Using LMConv, we learn an ensemble of distribution estimators that share parameters but differ in generation order, achieving improved performance on whole-image density estimation (2.89 bpd on unconditional CIFAR10), as well as globally coherent image completions. Our code is available at https://ajayjain.github.io/lmconv.

One fundamental problem in deep learning is understanding the outstanding performance of deep Neural Networks (NNs) in practice. One explanation for the superiority of NNs is that they can realize a large class of complicated functions, i.e., they have powerful expressivity. The expressivity of a ReLU NN can be quantified by the maximal number of linear regions it can separate its input space into. In this paper, we provide several mathematical results needed for studying the linear regions of CNNs, and use them to derive the maximal and average numbers of linear regions for one-layer ReLU CNNs. Furthermore, we obtain upper and lower bounds for the number of linear regions of multi-layer ReLU CNNs. Our results suggest that deeper CNNs have more powerful expressivity than their shallow counterparts, while CNNs have more expressivity than fully-connected NNs per parameter.

Graph Neural Networks (GNNs) are powerful tools in representation learning for graphs. However, recent studies show that GNNs are vulnerable to carefully-crafted perturbations, called adversarial attacks. Adversarial attacks can easily fool GNNs in making predictions for downstream tasks. The vulnerability to adversarial attacks has raised increasing concerns for applying GNNs in safety-critical applications. Therefore, developing robust algorithms to defend adversarial attacks is of great significance. A natural idea to defend adversarial attacks is to clean the perturbed graph. It is evident that real-world graphs share some intrinsic properties. For example, many real-world graphs are low-rank and sparse, and the features of two adjacent nodes tend to be similar. In fact, we find that adversarial attacks are likely to violate these graph properties. Therefore, in this paper, we explore these properties to defend adversarial attacks on graphs. In particular, we propose a general framework Pro-GNN, which can jointly learn a structural graph and a robust graph neural network model from the perturbed graph guided by these properties. Extensive experiments on real-world graphs demonstrate that the proposed framework achieves significantly better performance compared with the state-of-the-art defense methods, even when the graph is heavily perturbed. We release the implementation of Pro-GNN to our DeepRobust repository for adversarial attacks and defenses (footnote: https://github.com/DSE-MSU/DeepRobust). The specific experimental settings to reproduce our results can be found in https://github.com/ChandlerBang/Pro-GNN.