Reading comprehension models often overfit to nuances of training datasets and fail at adversarial evaluation. Training with adversarially augmented dataset improves robustness against those adversarial attacks but hurts generalization of the models. In this work, we present several effective adversaries and automated data augmentation policy search methods with the goal of making reading comprehension models more robust to adversarial evaluation, but also improving generalization to the source domain as well as new domains and languages. We first propose three new methods for generating QA adversaries, that introduce multiple points of confusion within the context, show dependence on insertion location of the distractor, and reveal the compounding effect of mixing adversarial strategies with syntactic and semantic paraphrasing methods. Next, we find that augmenting the training datasets with uniformly sampled adversaries improves robustness to the adversarial attacks but leads to decline in performance on the original unaugmented dataset. We address this issue via RL and more efficient Bayesian policy search methods for automatically learning the best augmentation policy combinations of the transformation probability for each adversary in a large search space. Using these learned policies, we show that adversarial training can lead to significant improvements in in-domain, out-of-domain, and cross-lingual (German, Russian, Turkish) generalization without any use of training data from the target domain or language.

The Intelligent GeoSolutions (IGS) team at the University of Maine's Center for Research on Sustainable Forests (CRSF) has released a free interactive mapping tool, the Forest Ecosystem Status a … Trends (ForEST) app, to provide online decision support to private and public forest managers, natural resource agencies, conservation organizations and other stakeholders. With the current outbreak of eastern spruce budworm expanding south from Quebec, up-to-date information about resource conditions and near-term risk are needed to coordinate mitigation actions in response to the outbreak and related market conditions. The ForEST app is the culmination of three years of research and software development by the IGS team in partnership with UMaine's Advanced Computing Group. The interdisciplinary project supported two graduate students in the School of Computing and Information Science, each of whom served as lead developer, as well as undergraduate computer science students who worked as team programmers. The interactive web interface is designed to provide near real-time information about changing forest landscape conditions resulting from the spruce budworm outbreak and ongoing management.

Intelligent tutoring systems have been shown to be effective in helping to teach certain subjects, such as algebra or grammar, but creating these computerized systems is difficult and laborious. Now, researchers at Carnegie Mellon University have shown they can rapidly build them by, in effect, teaching the computer to teach. Using a new method that employs artificial intelligence, a teacher can teach the computer by demonstrating several ways to solve problems in a topic, such as multicolumn addition, and correcting the computer if it responds incorrectly. Notably, the computer system learns to not only solve the problems in the ways it was taught, but also to generalize to solve all other problems in the topic, and do so in ways that might differ from those of the teacher, said Daniel Weitekamp III, a Ph.D. student in CMU's Human-Computer Interaction Institute (HCII). "A student might learn one way to do a problem and that would be sufficient," Weitekamp explained.

Python is one of the most preferred high-level programming languages, which is being increasingly utilised in data science and in designing complex machine learning algorithms. In one of our articles, we discussed why one should learn the Python programming language for data science and machine learning. In this article, we list down the top 9 free resources to learn Python for Machine Learning. About: This is a free class provided by the developers at Google. It includes written materials, lecture videos, and lots of code exercises to practice Python coding. The first exercises work on basic Python concepts like strings and lists, building up to the later exercises which are full programs dealing with text files, processes, and Http connections.

A proper grasp of statistics is essential for any machine learning enthusiast to succeed in the competitive domain. Consequently, one should focus more on statistics than on the latest fancy techniques. The book -- All of Statistics -- consists of 24 chapters and covers every topic right from probability to statistical inference and statistical models and methods.

Peloton has added a new groups feature to allow people to exercise together despite being in lockdown. The company – which makes internet-enabled spinning bikes and treadmills, as well as running an app of online exercise classes – has seen a huge surge in users in recent months as people have looked for a way to work out at home. Now it has added a new groups feature, officially called "tags", allowing people to track their exercise alongside other people in communities they have formed off the bike. With the new feature, users are able to add hashtags to their profile, which designate certain groups: a certain set of people all from the same workplace, for instance, or one of the many "tribes" of users that have formed on other platforms such as Facebook and Reddit. If a user has a given hashtag in their profile, they will be able to see what classes other members have taken and when they are working out.

Neural networks are known to suffer from catastrophic forgetting when trained on sequential datasets. While there have been numerous attempts to solve this problem for large-scale supervised classification, little has been done to overcome catastrophic forgetting for few-shot classification problems. We demonstrate that the popular gradient-based few-shot meta-learning algorithm Model-Agnostic Meta-Learning (MAML) indeed suffers from catastrophic forgetting and introduce a Bayesian online meta-learning framework that tackles this problem. Our framework incorporates MAML into a Bayesian online learning algorithm with Laplace approximation. This framework enables few-shot classification on a range of sequentially arriving datasets with a single meta-learned model. The experimental evaluations demonstrate that our framework can effectively prevent forgetting in various few-shot classification settings compared to applying MAML sequentially.

At LinkedIn, we want to create economic opportunity for everyone in the global workforce. A critical aspect of this goal is matching jobs with qualified applicants. To improve hiring efficiency and reduce the need to manually screening each applicant, we develop a new product where recruiters can ask screening questions online so that they can filter qualified candidates easily. To add screening questions to all $20$M active jobs at LinkedIn, we propose a new task that aims to automatically generate screening questions for a given job posting. To solve the task of generating screening questions, we develop a two-stage deep learning model called Job2Questions, where we apply a deep learning model to detect intent from the text description, and then rank the detected intents by their importance based on other contextual features. Since this is a new product with no historical data, we employ deep transfer learning to train complex models with limited training data. We launched the screening question product and our AI models to LinkedIn users and observed significant impact in the job marketplace. During our online A/B test, we observed $+53.10\%$ screening question suggestion acceptance rate, $+22.17\%$ job coverage, $+190\%$ recruiter-applicant interaction, and $+11$ Net Promoter Score. In sum, the deployed Job2Questions model helps recruiters to find qualified applicants and job seekers to find jobs they are qualified for.

In peer selection agents must choose a subset of themselves for an award or a prize. As agents are self-interested, we want to design algorithms that are impartial, so that an individual agent cannot affect their own chance of being selected. This problem has broad application in resource allocation and mechanism design and has received substantial attention in the artificial intelligence literature. Here, we present a novel algorithm for impartial peer selection, PeerNomination, and provide a theoretical analysis of its accuracy. Our algorithm possesses various desirable features. In particular, it does not require an explicit partitioning of the agents, as previous algorithms in the literature. We show empirically that it achieves higher accuracy than the exiting algorithms over several metrics.

Genetic Algorithms (GAs) are members of a general class of optimization algorithms, known as Evolutionary Algorithms (EAs), which simulate a fictional environment based on theory of evolution to deal with various types of mathematical problem, especially those related to optimization. Also Genetic Algorithms can be categorized as a subset of Metaheuristics, which are general-purpose tools and algorithms to solve optimization and unsupervised learning problems. In this series of video tutorials, we are going to learn about Genetic Algorithms, from theory to implementation. After having a brief review of theories behind EA and GA, two main versions of genetic algorithms, namely Binary Genetic Algorithm and Real-coded Genetic Algorithm, are implemented from scratch and line-by-line, using both Python and MATLAB. This course is instructed by Dr. Mostapha Kalami Heris, who has years of practical work and active teaching in the field of computational intelligence.