Semi-supervised learning (SSL) provides a powerful framework for leveraging unlabeled data when labels are limited or expensive to obtain. SSL algorithms based on deep neural networks have recently proven successful on standard benchmark tasks. However, we argue that these benchmarks fail to address many issues that these algorithms would face in real-world applications. After creating a unified reimplementation of various widely-used SSL techniques, we test them in a suite of experiments designed to address these issues. We find that the performance of simple baselines which do not use unlabeled data is often underreported, that SSL methods differ in sensitivity to the amount of labeled and unlabeled data, and that performance can degrade substantially when the unlabeled dataset contains out-of-class examples. To help guide SSL research towards real-world applicability, we make our unified reimplemention and evaluation platform publicly available.

Iterative thresholding algorithms seek to optimize a differentiable objective function over a sparsity or rank constraint by alternating between gradient steps that reduce the objective, and thresholding steps that enforce the constraint. This work examines the choice of the thresholding operator, and asks whether it is possible to achieve stronger guarantees than what is possible with hard thresholding. We develop the notion of relative concavity of a thresholding operator, a quantity that characterizes the convergence performance of any thresholding operator on the target optimization problem. Surprisingly, we find that commonly used thresholding operators, such as hard thresholding and soft thresholding, are suboptimal in terms of convergence guarantees. Instead, a general class of thresholding operators, lying between hard thresholding and soft thresholding, is shown to be optimal with the strongest possible convergence guarantee among all thresholding operators. Examples of this general class includes $\ell_q$ thresholding with appropriate choices of $q$, and a newly defined {\em reciprocal thresholding} operator. As a byproduct of the improved convergence guarantee, these new thresholding operators improve on the best known upper bound for prediction error of both iterative hard thresholding and Lasso in terms of the dependence on condition number in the setting of sparse linear regression.

Social media provides a mechanism for people to engage with social causes across a range of issues. It also provides a strategic tool to those looking to advance a cause to exchange, promote or publicize their ideas. In such instances, AI can be either an asset if used appropriately or a barrier. One of the key issues for a workforce diversity campaign is to understand in real-time who is participating - specifically, whether the participants are individuals or organizations, and in case of individuals, whether they are male or female. In this paper, we present a study to demonstrate a case for AI for social good that develops a model to infer in real-time the different user types participating in a cause-driven hashtag campaign on Twitter, ILookLikeAnEngineer (ILLAE). A generic framework is devised to classify a Twitter user into three classes: organization, male and female in a real-time manner. The framework is tested against two datasets (IL-LAE and a general dataset) and outperforms the baseline binary classifiers for categorizing organization/individual and male/female. The proposed model can be applied to future social cause-driven campaigns to get real-time insights on the macro-level social behavior of participants.

We propose a technique for declaratively specifying strategies for semi-supervised learning (SSL). The proposed method can be used to specify ensembles of semi-supervised learning, as well as agreement constraints and entropic regularization constraints between these learners, and can be used to model both well-known heuristics such as co-training and novel domain-specific heuristics. In addition to representing individual SSL heuristics, we show that multiple heuristics can also be automatically combined using Bayesian optimization methods. We show consistent improvements on a suite of well-studied SSL benchmarks, including a new state-of-the-art result on a difficult relation extraction task.

We build a virtual agent for learning language in a 2D maze-like world. The agent sees images of the surrounding environment, listens to a virtual teacher, and takes actions to receive rewards. It interactively learns the teacher's language from scratch based on two language use cases: sentence-directed navigation and question answering. It learns simultaneously the visual representations of the world, the language, and the action control. By disentangling language grounding from other computational routines and sharing a concept detection function between language grounding and prediction, the agent reliably interpolates and extrapolates to interpret sentences that contain new word combinations or new words missing from training sentences. The new words are transferred from the answers of language prediction. Such a language ability is trained and evaluated on a population of over 1.6 million distinct sentences consisting of 119 object words, 8 color words, 9 spatial-relation words, and 50 grammatical words. The proposed model significantly outperforms five comparison methods for interpreting zero-shot sentences. In addition, we demonstrate human-interpretable intermediate outputs of the model in the appendix.

Deep learning is well known to be very amenable to GPU acceleration. Accelerating "traditional" machine learning methods like logistic regression, linear regression, and support vector machines with GPUs at scale, has, however, been challenging. Today I am very proud to share a major breakthrough that IBM Research has made in this critical area. A team out of our Zurich IBM Research lab beat a previous performance benchmark set for a machine learning workload by Google by 46 times. The research team trained a logistic regression classifier to predict clicks on advertisements using a Terabyte-scale data set that consists of online advertising click-thru data, containing 4.2 billion training examples and 1 million features.

Amazon Sagemaker was launched by Amazon back in November 2017. It was built with the promise of simplifying machine learning on the cloud. The software was a response not only to the increasing importance of machine learning, but also the fact that there is a demand to perform machine learning in the cloud. Amazon Sagemaker is clearly a smart move by Amazon that will consolidate the dominance of AWS in the cloud market. Amazon Sagemaker is Amazon's premium cloud-based service which serves as a platform for machine learning developers and data scientists to build, train and deploy machine learning models on the cloud.

The purpose of this paper is to re-investigate the estimation of multiple factor models by relaxing the convention that the number of factors is small. We first obtain the collection of all possible factors and we provide a simultaneous test, security by security, of which factors are significant. Since the collection of risk factors selected for investigation is large and highly correlated, we use dimension reduction methods, including the Least Absolute Shrinkage and Selection Operator (LASSO) and prototype clustering, to perform the investigation. For comparison with the existing literature, we compare the multi-factor model's performance with the Fama-French 5-factor model. We find that both the Fama-French 5-factor and the multi-factor model are consistent with the behavior of "large-time scale" security returns. In a goodness-of-fit test comparing the Fama-French 5-factor with the multi-factor model, the multi-factor model has a substantially larger adjusted $R^{2}$. Robustness tests confirm that the multi-factor model provides a reasonable characterization of security returns.

Stochastic optimization is key to efficient inversion in PDE-constrained optimization. Using 'simultaneous shots', or random superposition of source terms, works very well in simple acquisition geometries where all sources see all receivers, but this rarely occurs in practice. We develop an approach that interpolates data to an ideal acquisition geometry while solving the inverse problem using simultaneous shots. The approach is formulated as a joint inverse problem, combining ideas from low-rank interpolation with full-waveform inversion. Results using synthetic experiments illustrate the flexibility and efficiency of the approach.

This work adopts the very successful distributional perspective on reinforcement learning and adapts it to the continuous control setting. We combine this within a distributed framework for off-policy learning in order to develop what we call the Distributed Distributional Deep Deterministic Policy Gradient algorithm, D4PG. We also combine this technique with a number of additional, simple improvements such as the use of $N$-step returns and prioritized experience replay. Experimentally we examine the contribution of each of these individual components, and show how they interact, as well as their combined contributions. Our results show that across a wide variety of simple control tasks, difficult manipulation tasks, and a set of hard obstacle-based locomotion tasks the D4PG algorithm achieves state of the art performance.