We investigate the problem of truth discovery based on opinions from multiple agents who may be unreliable or biased. We consider the case where agents' reliabilities or biases are correlated if they belong to the same community, which defines a group of agents with similar opinions regarding a particular event. An agent can belong to different communities for different events, and these communities are unknown \emph{a priori}. We incorporate knowledge of the agents' social network in our truth discovery framework and develop Laplace variational inference methods to estimate agents' reliabilities, communities, and the event states. We also develop a stochastic variational inference method to scale our model to large social networks. Simulations and experiments on real data suggest that when observations are sparse, our proposed methods perform better than several other inference methods, including majority voting, the popular Bayesian Classifier Combination (BCC) method, and the Community BCC method.

We develop a novel framework for estimating causal effects based on the discrepancy between unobserved counterfactual distributions. In our setting a causal effect is defined in terms of the $L_1$ distance between different counterfactual outcome distributions, rather than a mean difference in outcome values. Directly comparing counterfactual outcome distributions can provide more nuanced and valuable information about causality than a simple comparison of means. We consider single- and multi-source randomized studies, as well as observational studies, and analyze error bounds and asymptotic properties of the proposed estimators. We further propose methods to construct confidence intervals for the unknown mean distribution distance. Finally, we illustrate the new methods and verify their effectiveness in empirical studies.

Many structured prediction problems (particularly in vision and language domains) are ambiguous, with multiple outputs being correct for an input - e.g. there are many ways of describing an image, multiple ways of translating a sentence; however, exhaustively annotating the applicability of all possible outputs is intractable due to exponentially large output spaces (e.g. all English sentences). In practice, these problems are cast as multi-class prediction, with the likelihood of only a sparse set of annotations being maximized - unfortunately penalizing for placing beliefs on plausible but unannotated outputs. We make and test the following hypothesis - for a given input, the annotations of its neighbors may serve as an additional supervisory signal. Specifically, we propose an objective that transfers supervision from neighboring examples. We first study the properties of our developed method in a controlled toy setup before reporting results on multi-label classification and two image-grounded sequence modeling tasks - captioning and question generation. We evaluate using standard task-specific metrics and measures of output diversity, finding consistent improvements over standard maximum likelihood training and other baselines.

Meta-learning for few-shot learning entails acquiring a prior over previous tasks and experiences, such that new tasks be learned from small amounts of data. However, a critical challenge in few-shot learning is task ambiguity: even when a powerful prior can be meta-learned from a large number of prior tasks, a small dataset for a new task can simply be too ambiguous to acquire a single model (e.g., a classifier) for that task that is accurate. In this paper, we propose a probabilistic meta-learning algorithm that can sample models for a new task from a model distribution. Our approach extends model-agnostic meta-learning, which adapts to new tasks via gradient descent, to incorporate a parameter distribution that is trained via a variational lower bound. At meta-test time, our algorithm adapts via a simple procedure that injects noise into gradient descent, and at meta-training time, the model is trained such that this stochastic adaptation procedure produces samples from the approximate model posterior. Our experimental results show that our method can sample plausible classifiers and regressors in ambiguous few-shot learning problems.

This paper explores the use of adversarial examples in training speech recognition systems to increase robustness of deep neural network acoustic models. During training, the fast gradient sign method is used to generate adversarial examples augmenting the original training data. Different from conventional data augmentation based on data transformations, the examples are dynamically generated based on current acoustic model parameters. We assess the impact of adversarial data augmentation in experiments on the Aurora-4 and CHiME-4 single-channel tasks, showing improved robustness against noise and channel variation. Further improvement is obtained when combining adversarial examples with teacher/student training, leading to a 23% relative word error rate reduction on Aurora-4.

Get your team access to Udemy's top 2,500 courses anytime, anywhere.

With the 19th International Conference on Artificial Intelligence in Education (AIED) being held later this month, it might be a good time to start thinking about the very possible future of AI in schools and what it might mean for your children -- both the good and the bad. If you're a parent or in the education field, you will no doubt be up to date on the ways that education in public schools and universities across the nation have already been steadily adapting towards a more technology-driven and data-producing form. Gone are the dusty chalkboards, the endless piles of handwritten essays, the class trips to the magical and always-too-warm computer lab, and those annoying moments when "that" student decides to spend 25 minutes staring out of the window while "using the pencil sharpener." Instead, students across the nation are utilizing Google's Chromebooks and collaborating on shared documents. What's more, they're learning to code as a basic element of their curriculum.

Summary: There is a great hue and cry about the danger of bias in our predictive models when applied to high significance events like who gets a loan, insurance, a good school assignment, or bail. It's not as simple as it seems and here we try to take a more nuanced look. The result is not as threatening as many headlines make it seem. Is social bias in our models a threat to equal and fair treatment? There's even an entire conference dedicated to the topic: the conference on Fairness, Accountability, and Transparency (FAT* – it's their acronym, I didn't make this up) now in its fifth year.

The corporate training market is over $200 billion around the world[1] and it's going through a revolution. While we often think of training as programs or courses, a new paradigm has arrived, one I call "Learning in the Flow of Work." The corporate training industry has been around for decades and it has always been impacted by new technology. As the following chart shows, over the last 20 years we've been through four evolutions, each driven by technological and economic change. In the 1970s and 1980s, when I started my career, we learned in classrooms. The technology was slide projectors and "foils" (plastic laminated slides).

Databricks today unveiled MLflow, a new open source project that aims to provide some standardization to the complex processes that data scientists oversee during the course of building, testing, and deploying machine learning models. "Everybody who has done machine learning knows that the machine learning development lifecycle is very complex," Apache Spark creator and Databricks CTO Matei Zaharia said during his keynote address at Databricks' Spark and AI Summit in San Francisco. "There are a lot of issues that come up that you don't have in normal software development lifecycle." The vast volumes of data, together with the abundance of machine learning frameworks, the large scale of production systems, and the distributed nature of data science and engineering teams, combine to provide a huge number of variables to control in the machine learning DevOps lifecycle -- and that even before the tuning. "They have all these tuning parameters that you have to change and explore to get a good model," Zaharia said.