Adversarial sample attacks perturb benign inputs to induce DNN misbehaviors. Recent research has demonstrated the widespread presence and the devastating consequences of such attacks. Existing defense techniques either assume prior knowledge of specific attacks or may not work well on complex models due to their underlying assumptions. We argue that adversarial sample attacks are deeply entangled with interpretability of DNN models: while classification results on benign inputs can be reasoned based on the human perceptible features/attributes, results on adversarial samples can hardly be explained. Therefore, we propose a novel adversarial sample detection technique for face recognition models, based on interpretability. It features a novel bi-directional correspondence inference between attributes and internal neurons to identify neurons critical for individual attributes. The activation values of critical neurons are enhanced to amplify the reasoning part of the computation and the values of other neurons are weakened to suppress the uninterpretable part. The classification results after such transformation are compared with those of the original model to detect adversaries. Results show that our technique can achieve 94% detection accuracy for 7 different kinds of attacks with 9.91% false positives on benign inputs. In contrast, a state-of-the-art feature squeezing technique can only achieve 55% accuracy with 23.3% false positives.

This paper proposes a framework based on deep convolutional neural networks (CNNs) for automatic heart sound classification using short-segments of individual heart beats. We design a 1D-CNN that directly learns features from raw heart-sound signals, and a 2D-CNN that takes inputs of two- dimensional time-frequency feature maps based on Mel-frequency cepstral coefficients (MFCC). We further develop a time-frequency CNN ensemble (TF-ECNN) combining the 1D-CNN and 2D-CNN based on score-level fusion of the class probabilities. On the large PhysioNet CinC challenge 2016 database, the proposed CNN models outperformed traditional classifiers based on support vector machine and hidden Markov models with various hand-crafted time- and frequency-domain features. Best classification scores with 89.22% accuracy and 89.94% sensitivity were achieved by the ECNN, and 91.55% specificity and 88.82% modified accuracy by the 2D-CNN alone on the test set.

We introduce an extension of the multi-instance learning problem where examples are organized as nested bags of instances (e.g., a document could be represented as a bag of sentences, which in turn are bags of words). This framework can be useful in various scenarios, such as text and image classification, but also supervised learning over graphs. As a further advantage, multi-multi instance learning enables a particular way of interpreting predictions and the decision function. Our approach is based on a special neural network layer, called bag-layer, whose units aggregate bags of inputs of arbitrary size. We prove theoretically that the associated class of functions contains all Boolean functions over sets of sets of instances and we provide empirical evidence that functions of this kind can be actually learned on semi-synthetic datasets. We finally present experiments on text classification and on citation graphs and social graph data, showing that our model obtains competitive results with respect to other approaches such as convolutional networks on graphs.

ABSTRACT Typical spoken language understanding systems provide narrow semantic parses using a domain-specific ontology. The parses contain intents and slots that are directly consumed by downstream domain applications. In this work we discuss expanding such systems to handle compound entities and intents by introducing a domain-agnostic shallow parser that handles linguistic coordination. We show that our model for parsing coordination learns domain-independent and slot-independent features and is able to segment conjunct boundaries of many different phrasal categories. We also show that using adversarial training can be effective for improving generalization across different slot types for coordination parsing. Index Terms-- spoken language understanding, chunking, coordination 1. INTRODUCTION A typical spoken language understanding (SLU) system maps user utterances to domain-specific semantic representations that can be factored into an intent and slots [1, 2]. For example, an utterance, "what is the weather like in boston" has one intent WeatherInfo and one slot type CityName whose value is "boston." Thus, parsing for such systems is often factored into two separate tasks: intent classification and entity recognition whose results are consumed by downstream domain applications.

We propose Learned Accept/Reject Sampling (Lars), a method for constructing richer priors using rejection sampling with a learned acceptance function. This work is motivated by recent analyses of the VAE objective, which pointed out that commonly used simple priors can lead to underfitting. As the distribution induced by Lars involves an intractable normalizing constant, we show how to estimate it and its gradients efficiently. We demonstrate that Lars priors improve VAE performance on several standard datasets both when they are learned jointly with the rest of the model and when they are fitted to a pretrained model. Finally, we show that Lars can be combined with existing methods for defining flexible priors for an additional boost in performance.

Discovery of atomistic systems with desirable properties is a major challenge in chemistry and material science. Here we introduce a novel, autoregressive, convolutional deep neural network architecture that generates molecular equilibrium structures by sequentially placing atoms in three-dimensional space. The model estimates the joint probability over molecular configurations with tractable conditional probabilities which only depend on distances between atoms and their nuclear charges. It combines concepts from state-of-the-art atomistic neural networks with auto-regressive generative models for images and speech. We demonstrate that the architecture is capable of generating molecules close to equilibrium for constitutional isomers of C$_7$O$_2$H$_{10}$.

This paper considers the problem of recovering signals from compressed measurements contaminated with sparse outliers, which has arisen in many applications. In this paper, we propose a generative model neural network approach for reconstructing the ground truth signals under sparse outliers. We propose an iterative alternating direction method of multipliers (ADMM) algorithm for solving the outlier detection problem via $\ell_1$ norm minimization, and a gradient descent algorithm for solving the outlier detection problem via squared $\ell_1$ norm minimization. We establish the recovery guarantees for reconstruction of signals using generative models in the presence of outliers, and give an upper bound on the number of outliers allowed for recovery. Our results are applicable to both the linear generator neural network and the nonlinear generator neural network with an arbitrary number of layers. We conduct extensive experiments using variational auto-encoder and deep convolutional generative adversarial networks, and the experimental results show that the signals can be successfully reconstructed under outliers using our approach. Our approach outperforms the traditional Lasso and $\ell_2$ minimization approach.

We propose a real-time context-aware learning system along with the architecture that runs on the mobile devices, provide services to the user and manage the IoT devices. In this system, an application running on mobile devices collected data from the sensors, learned about the user-defined context, made predictions in real-time and manage IoT devices accordingly. However, the computational power of the mobile devices makes it challenging to run machine learning algorithms with acceptable accuracy. To solve this issue, some authors have run machine learning algorithms on the server and transmitted the results to the mobile devices. Although the context-aware predictions made by the server are more accurate than their mobile counterpart, it heavily depends on the network connection for the delivery of the results to the devices, which negatively affects real-time context-learning. Therefore, in this work, we describe a context-learning algorithm for mobile devices which is less demanding on the computational resources and maintains the accuracy of the prediction by updating itself from the learning parameters obtained from the server periodically. Corresponding author Email address: das.bhaskar.1981@gmail.com, Keywords: Mobile computing, Context-aware Applications, Real-time System, Context learning, Cloud Computing 1. Introduction The Internet of Things (IoT) is a vision where everyday devices embedded with computing technology can communicate with one another via the Internet. The IoT devices will play an important role in improving quality of life in many domains such as smart living, transportation, education, agriculture, industry, and the like. By 2020, approximately 212 billion devices will be deployed globally [1] and will consume 45% of the internet traffic by 2022 [2]. IoT-related healthcare industry is expected to grow to $1.1 - $2.5 trillion by 2025 globally [3], and according to "Navigant Research", the building automation systems market is expected to reach $102.0 billion in 2025. Due to the technological advancement of the mobile devices, a new application domain has emerged, called context-aware computing, in which the system can make use of environmental information from collected sensor data and respond accordingly without requiring any user intervention [5].

In laser microsurgery, image-based control of the ablation laser can lead to higher accuracy and patient safety. However, camera-based image acquisition lacks the subcutaneous tissue perception. Optical coherence tomography (OCT) as high-resolution imaging modality yields transsectional images of tissue and can provide the missing depth information. Therefore, this paper deals with the tracking of distinctive subcutaneous structures on OCTs for automated control of ablation lasers in microsurgery. We present a deep learning based tracking scheme for concise representations of subsequent 3D OCT volumes. For each of the volume, a compact representation is created by calculating maximum intensity projections and projecting the depth value, were the maximum intensity voxel is found, onto an image plane. These depth images are then used for tracking by estimating the dense optical flow and depth changes with a self-supervisely trained convolutional neural network. Tracking performances are evaluated on a dataset of ex vivo human temporal bone with rigid ground truth transformations and on an in vivo sequence of human skin with non-rigid transformations. First quantitative evaluation reveals a mean endpoint error of 2.27voxel for scene flow estimation. Object tracking on 4D OCT data enables its use for sub-epithelial tracking of tissue structures for image-guidance in automated laser incision control for microsurgery.

We propose a method to perform audio event detection under the common constraint that only limited training data are available. In training a deep learning system to perform audio event detection, two practical problems arise. Firstly, most datasets are "weakly labelled" having only a list of events present in each recording without any temporal information for training. Secondly, deep neural networks need a very large amount of labelled training data to achieve good quality performance, yet in practice it is difficult to collect enough samples for most classes of interest. In this paper, we propose a data-efficient training of a stacked convolutional and recurrent neural network. This neural network is trained in a multi instance learning setting for which we introduce a new loss function that leads to improved training compared to the usual approaches for weakly supervised learning. We successfully test our approach on two low-resource datasets that lack temporal labels.