Survival analysis has been developed and applied in the number of areas including manufacturing, finance, economics and healthcare. In healthcare domain, usually clinical data are high-dimensional, sparse and complex and sometimes there exists few amount of time-to-event (labeled) instances. Therefore building an accurate survival model from electronic health records is challenging. With this motivation, we address this issue and provide a new survival analysis framework using deep learning and active learning with a novel sampling strategy. First, our approach provides better representation with lower dimensions from clinical features using labeled (time-to-event) and unlabeled (censored) instances and then actively trains the survival model by labeling the censored data using an oracle. As a clinical assistive tool, we introduce a simple effective treatment recommendation approach based on our survival model. In the experimental study, we apply our approach on SEER-Medicare data related to prostate cancer among African-Americans and white patients. The results indicate that our approach outperforms significantly than baseline models.

Deep generative models have achieved great success in unsupervised learning with the ability to capture complex nonlinear relationships between latent generating factors and observations. Among them, a factorized hierarchical variational autoencoder (FHVAE) is a variational inference-based model that formulates a hierarchical generative process for sequential data. Specifically, an FHVAE model can learn disentangled and interpretable representations, which have been proven useful for numerous speech applications, such as speaker verification, robust speech recognition, and voice conversion. However, as we will elaborate in this paper, the training algorithm proposed in the original paper is not scalable to datasets of thousands of hours, which makes this model less applicable on a larger scale. After identifying limitations in terms of runtime, memory, and hyperparameter optimization, we propose a hierarchical sampling training algorithm to address all three issues. Our proposed method is evaluated comprehensively on a wide variety of datasets, ranging from 3 to 1,000 hours and involving different types of generating factors, such as recording conditions and noise types. In addition, we also present a new visualization method for qualitatively evaluating the performance with respect to interpretability and disentanglement. Models trained with our proposed algorithm demonstrate the desired characteristics on all the datasets.

The field of artificial neural networks has exploded during the 1980s due to its universal approximation capabilities, as can be seen in [1], but the lack of understanding of the underlying statistical and geometric features extracted from the analyzed signal discouraged significantly its usage among scientists and researchers, as can be seen in [2-3]. Since then, most of its usage has been relegated to applications where such understanding can be neglected, such as computer vision, nonlinear statespace estimators and other tasks related to control where exact algorithmic approaches are unknown or too difficult to implement, according to [3]. More recently, aiming to enlightening these black-boxes, several approaches have been under heavy development, such as variables contributions in the feed forward structure [4], visualization using saliency maps [5], generation of skeletal structures [6], fuzzy rule based evaluation of all permutations [3], extraction of functional relations using sensitivity analysis of input data [7], as many others. In a parallel way, other researchers have been successfully developing new kinds of feed-forward neural architectures that behave much more like a transparent box, where the extracted features can be directly evaluated and understood. Convolutional Neural Networks are a great example of such achievements, as can be seen in [8-10]. Despite its several layers, they can be employed on different types of tasks, including text classification, natural language processing, computer vision and so on, with a good understanding of what is happening behind the curtains. Manuscript received January 15, 2018. This work was supported in part by the FIPE (Institute of Economic Research Foundation) by means of a postdoctoral scholarship.

Techniques such as ensembling and distillation promise model quality improvements when paired with almost any base model. However, due to increased test-time cost (for ensembles) and increased complexity of the training pipeline (for distillation), these techniques are challenging to use in industrial settings. In this paper we explore a variant of distillation which is relatively straightforward to use as it does not require a complicated multi-stage setup or many new hyperparameters. Our first claim is that online distillation enables us to use extra parallelism to fit very large datasets about twice as fast. Crucially, we can still speed up training even after we have already reached the point at which additional parallelism provides no benefit for synchronous or asynchronous stochastic gradient descent. Two neural networks trained on disjoint subsets of the data can share knowledge by encouraging each model to agree with the predictions the other model would have made. These predictions can come from a stale version of the other model so they can be safely computed using weights that only rarely get transmitted. Our second claim is that online distillation is a cost-effective way to make the exact predictions of a model dramatically more reproducible. We support our claims using experiments on the Criteo Display Ad Challenge dataset, ImageNet, and the largest to-date dataset used for neural language modeling, containing $6\times 10^{11}$ tokens and based on the Common Crawl repository of web data.

Deep neural networks (DNNs) are known vulnerable to adversarial attacks. That is, adversarial examples, obtained by adding delicately crafted distortions onto original legal inputs, can mislead a DNN to classify them as any target labels. In a successful adversarial attack, the targeted mis-classification should be achieved with the minimal distortion added. In the literature, the added distortions are usually measured by L0, L1, L2, and L infinity norms, namely, L0, L1, L2, and L infinity attacks, respectively. However, there lacks a versatile framework for all types of adversarial attacks. This work for the first time unifies the methods of generating adversarial examples by leveraging ADMM (Alternating Direction Method of Multipliers), an operator splitting optimization approach, such that L0, L1, L2, and L infinity attacks can be effectively implemented by this general framework with little modifications. Comparing with the state-of-the-art attacks in each category, our ADMM-based attacks are so far the strongest, achieving both the 100% attack success rate and the minimal distortion.

Modern health data science applications leverage abundant molecular and electronic health data, providing opportunities for machine learning to build statistical models to support clinical practice. Time-to-event analysis, also called survival analysis, stands as one of the most representative examples of such statistical models. We present a novel deep-network-based approach that leverages adversarial learning to address a key challenge in modern time-to-event modeling: nonparametric estimation of event-time distributions. We also introduce a principled cost function to exploit information from censored events (events that occur subsequent to the observation window). Unlike most time-to-event models, we focus on the estimation of time-to-event distributions, rather than time ordering. We validate our model on both benchmark and real datasets, demonstrating that the proposed formulation yields significant performance gains relative to a parametric alternative, which we also propose.

Quantifying behavior is crucial for many applications in neuroscience. Videography provides easy methods for the observation and recording of animal behavior in diverse settings, yet extracting particular aspects of a behavior for further analysis can be highly time consuming. In motor control studies, humans or other animals are often marked with reflective markers to assist with computer-based tracking, yet markers are intrusive (especially for smaller animals), and the number and location of the markers must be determined a priori. Here, we present a highly efficient method for markerless tracking based on transfer learning with deep neural networks that achieves excellent results with minimal training data. We demonstrate the versatility of this framework by tracking various body parts in a broad collection of experimental settings: mice odor trail-tracking, egg-laying behavior in drosophila, and mouse hand articulation in a skilled forelimb task. For example, during the skilled reaching behavior, individual joints can be automatically tracked (and a confidence score is reported). Remarkably, even when a small number of frames are labeled ($\approx 200$), the algorithm achieves excellent tracking performance on test frames that is comparable to human accuracy.

Using software to parse the world's visual content is as big of a revolution in computing as mobile was 10 years ago, and will provide a major edge for developers and businesses to build amazing products. Computer Vision is the process of using machines to understand and analyze imagery (both photos and videos). While these types of algorithms have been around in various forms since the 1960's, recent advances in Machine Learning, as well as leaps forward in data storage, computing capabilities, and cheap high-quality input devices, have driven major improvements in how well our software can explore this kind of content. Computer Vision is the broad parent name for any computations involving visual content – that means images, videos, icons, and anything else with pixels involved. A classical application of computer vision is handwriting recognition for digitizing handwritten content (we'll explore more use cases below). Any other application that involves understanding pixels through software can safely be labeled as computer vision.

Helping us all make sense of, and enrich data that is moving along via our data pipes. It is common for our customers to perform sentiment analysis, enrich with tags, and extract names, dates, emails, and other relevant information for streams as they arrive, or as they are being delivered to other destinations. By adding additional tags, meaning, and other metadata, it makes it easier to connect and aggregate data across real-time streams, and transform existing streams into richer topical feeds. We are working on profiling, not just Algorithmia, but a number of other machine learning APIs. As we establish interesting collections of text analysis, deep learning, and other algorithms that can be applied to Streamdata.io streams, we'll publish here on the blog. If you have specific data and content, or machine learning model that you'd like to have delivered as part of your real-time infrastructure let us know. We are happy to prioritize specific types of data or profile more relevant machine learning APIs providers to help expedite your work. We are beginning to ramp up our efforts to profile relevant machine learning models, as the demand from our customers' increases, hoping to satisfy our customers demand for machine learning intelligence as they continue to optimize their streams of data across their organization.

In this contest, participants try to create the best agent for playing custom levels of the Sonic games -- without having access to those levels during development. See our blog post for more details. This process is illustrated in the schematic below. We believe that the next step for reinforcement learning is to leverage past experience to quickly learn new environments. Current algorithms are very prone to memorization and can't adapt well to new situations.