The vulnerability of deep neural networks to adversarial attacks has been widely demonstrated (e.g., adversarial example attacks). Traditional attacks perform unstructured pixel-wise perturbation to fool the classifier. An alternative approach is to have perturbations in the latent space. However, such perturbations are hard to control due to the lack of interpretability and disentanglement. In this paper, we propose a more practical adversarial attack by designing structured perturbation with semantic meanings. Our proposed technique manipulates the semantic attributes of images via the disentangled latent codes. The intuition behind our technique is that images in similar domains have some commonly shared but theme-independent semantic attributes, e.g. thickness of lines in handwritten digits, that can be bidirectionally mapped to disentangled latent codes. We generate adversarial perturbation by manipulating a single or a combination of these latent codes and propose two unsupervised semantic manipulation approaches: vector-based disentangled representation and feature map-based disentangled representation, in terms of the complexity of the latent codes and smoothness of the reconstructed images. We conduct extensive experimental evaluations on real-world image data to demonstrate the power of our attacks for black-box classifiers. We further demonstrate the existence of a universal, image-agnostic semantic adversarial example.

Natural language programming is a promising approach to enable end users to instruct new tasks for intelligent agents. However, our formative study found that end users would often use unclear, ambiguous or vague concepts when naturally instructing tasks in natural language, especially when specifying conditionals. Existing systems have limited support for letting the user teach agents new concepts or explaining unclear concepts. In this paper, we describe a new multi-modal domain-independent approach that combines natural language programming and programming-by-demonstration to allow users to first naturally describe tasks and associated conditions at a high level, and then collaborate with the agent to recursively resolve any ambiguities or vagueness through conversations and demonstrations. Users can also define new procedures and concepts by demonstrating and referring to contents within GUIs of existing mobile apps. We demonstrate this approach in PUMICE, an end-user programmable agent that implements this approach. A lab study with 10 users showed its usability.

Detecting individual activity on smartphones still seems to be a challenge given the limitations of resources such as battery life and computational workload capacity. Considering user activity and managing them, we can conceive low power consumption for mobile phones and other mobile devices, which requires a complete and rigorous program to recognize a ctivities and adjust device power consumption regarding their application at different times and places. However, with the rapid development of new and innovative applications for mobile devices such as smartphones, advances in battery technology do not ke ep up, especially in energy conservation. On the other hand, the use of activity recognition is increasing in active and preventive healthcare applications at home, learning environments of security systems, and a variety of human - computer interactions. Th is paper proposes and implements a system for activity recognition in the home environment with a set of switch sensors and a practical text - based sampling tool.

Language Models Are An Effective Patient Representation Learning Technique For Electronic Health Record Data Ethan Steinberg, Ken Jung, Jason A. Fries, Conor K. Corbin, Stephen R. Pfohl, Nigam H. Shah January 16, 2020 Abstract Widespread adoption of electronic health records (EHRs) has fueled development of clinical outcome models using machine learning. However, patient EHR data are complex, and how to optimally represent them is an open question. This complexity, along with often small training set sizes available to train these clinical outcome models, are two core challenges for training high quality models. In this paper, we demonstrate that learning generic representations from the data of all the patients in the EHR enables better performing prediction models for clinical outcomes, allowing for these challenges to be overcome. We adapt common representation learning techniques used in other domains and find that representations inspired by language models enable a 3.5% mean improvement in AUROC on five clinical outcomes compared to standard baselines, with the average improvement rising to 19% when only a small number of patients are available for training a prediction model for a given clinical outcome. 1 Introduction The widespread adoption of electronic health records (EHRs) has created opportunities for using machine learning to reduce healthcare costs and improve quality of care. EHR data have been used to learn prediction models for clinical outcomes such as mortality [1], sepsis [2], 30-day readmission [3] and others [4, 5]. However, the complexity of patient data poses many obstacles to its effective use. Patient records in EHRs are variable length, high dimensional and sparse, with complex temporal and hierarchical structure. They are comprised of irregularly spaced visits spread across years, with each visit consisting of a subset of thousands of possible diagnosis, procedure, and medication codes as well as lab values and unstructured data such as text or images. In contrast, most off-the-shelf machine learning algorithms expect a fixed length vector of features as input. Manually defining a transformation of patient records into such a representation beyond simple binned counts is time consuming and outcome-dependent, leaving much of the temporal and hierarchical structure of EHRs underutilized when building machine learning models. The challenge of representing EHR data can be addressed by using neural networks to automatically learn how to featurize patient data while learning a model for a given clinical outcome (e.g., mortality or 30 day readmissions) [4].

An approach to utilize recent advances in deep generative models for anomaly detection in a granular (continuous) sense on a real-world image dataset with quality issues is detailed using recent normalizing flow models, with implications in many other applications/domains/data types. The approach is completely unsupervised (no annotations available) but qualitatively shown to provide accurate semantic labeling for images via heatmaps of the scaled log-likelihood overlaid on the images. When sorted based on the median values per image, clear trends in quality are observed. Furthermore, downstream classification is shown to be possible and effective via a weakly supervised approach using the log-likelihood output from a normalizing flow model as a training signal for a feature-extracting convolutional neural network. The pre-linear dense layer outputs on the CNN are shown to disentangle high level representations and efficiently cluster various quality issues. Thus, an entirely non-annotated (fully unsupervised) approach is shown possible for accurate estimation and classification of quality issues..

Anomaly detection refers to the task of finding unusual instances that stand out from the normal data. In several applications, these outliers or anomalous instances are of greater interest compared to the normal ones. Specifically in the case of industrial optical inspection and infrastructure asset management, finding these defects (anomalous regions) is of extreme importance. Traditionally and even today this process has been carried out manually. Humans rely on the saliency of the defects in comparison to the normal texture to detect the defects. However, manual inspection is slow, tedious, subjective and susceptible to human biases. Therefore, the automation of defect detection is desirable. But for defect detection lack of availability of a large number of anomalous instances and labelled data is a problem. In this paper, we present a convolutional auto-encoder architecture for anomaly detection that is trained only on the defect-free (normal) instances. For the test images, residual masks that are obtained by subtracting the original image from the auto-encoder output are thresholded to obtain the defect segmentation masks. The approach was tested on two data-sets and achieved an impressive average F1 score of 0.885. The network learnt to detect the actual shape of the defects even though no defected images were used during the training.

Identifying the underlying directional relations from observational time series with nonlinear interactions and complex relational structures is key to a wide range of applications, yet remains a hard problem. In this work, we introduce a novel minimum predictive information regularization method to infer directional relations from time series, allowing deep learning models to discover nonlinear relations. Our method substantially outperforms other methods for learning nonlinear relations in synthetic datasets, and discovers the directional relations in a video game environment and a heart-rate vs. breath-rate dataset.

In the Information Bottleneck (IB), when tuning the relative strength between compression and prediction terms, how do the two terms behave, and what's their relationship with the dataset and the learned representation? In this paper, we set out to answer these questions by studying multiple phase transitions in the IB objective: $\text{IB}_\beta[p(z|x)] = I(X; Z) - \beta I(Y; Z)$ defined on the encoding distribution p(z|x) for input $X$, target $Y$ and representation $Z$, where sudden jumps of $dI(Y; Z)/d \beta$ and prediction accuracy are observed with increasing $\beta$. We introduce a definition for IB phase transitions as a qualitative change of the IB loss landscape, and show that the transitions correspond to the onset of learning new classes. Using second-order calculus of variations, we derive a formula that provides a practical condition for IB phase transitions, and draw its connection with the Fisher information matrix for parameterized models. We provide two perspectives to understand the formula, revealing that each IB phase transition is finding a component of maximum (nonlinear) correlation between $X$ and $Y$ orthogonal to the learned representation, in close analogy with canonical-correlation analysis (CCA) in linear settings. Based on the theory, we present an algorithm for discovering phase transition points. Finally, we verify that our theory and algorithm accurately predict phase transitions in categorical datasets, predict the onset of learning new classes and class difficulty in MNIST, and predict prominent phase transitions in CIFAR10.

Machine learning (ML) which was initially adopted for search ranking and recommendation systems has firmly moved into the realm of core enterprise operations like sales optimization and preventative healthcare. For such ML applications, often deployed in regulated environments, the standards for user privacy, security, and data governance are substantially higher. This imposes the need for tracking provenance end-to-end, from the data sources used for training ML models to the predictions of the deployed models. In this work, we take a first step towards this direction by introducing the ML provenance tracking problem in the context of data science scripts. The fundamental idea is to automatically identify the relationships between data and ML models and in particular, to track which columns in a dataset have been used to derive the features of a ML model. We discuss the challenges in capturing such provenance information in the context of Python, the most common language used by data scientists. We then, present Vamsa, a modular system that extracts provenance from Python scripts without requiring any changes to the user's code. Using up to 450K real-world data science scripts from Kaggle and publicly available Python notebooks, we verify the effectiveness of Vamsa in terms of coverage, and performance. We also evaluate Vamsa's accuracy on a smaller subset of manually labeled data. Our analysis shows that Vamsa's precision and recall range from 87.5% to 98.3% and its latency is typically in the order of milliseconds for scripts of average size.

--Training a neural network model can be a lifelong learning process and is a computationally intensive one. A severe adverse effect that may occur in deep neural network models is that they can suffer from catastrophic forgetting during retraining on new data. T o avoid such disruptions in the continuous learning, one appealing property is the additive nature of ensemble models. In this paper, we propose two generic ensemble approaches, gradient boosting and meta-learning, to solve the catastrophic forgetting problem in tuning pre-trained neural network models. With stationary training resources and various advanced neural network structures, deep learning models have exceeded human performance in many areas. However, a well-known limitation of deep learning models is the so-called "catastrophic forgetting."