Census data that shows that 231 million Americans speak only English at home and do not know another language well enough to communicate in it. But how can you learn a new language without going back to school? Machine learning could be a solution to this problem, by cutting down on the 200 hours it takes to learn a language using traditional methods. Language company Lingvist intends to decrease this time by using machine learning software to adapt to your learning style. The algorithm certainly seems to work well -- and the way certain words are reinforced makes sure that they stick in your mind.

Google this week started rolling out a revamped version of a mobile learning app, called Socratic, that the tech giant acquired last year. The updated app, with new machine learning-powered features, coincides with the start of the school year, as well as other Google for Education initiatives. Socratic aims to help both high school and university students in their studies outside of the classroom. If students need help answering a study question, they can now use the Socratic app to ask a question with their voice, or to take a picture of a question in their study materials. The app will then find relevant material from across the web.

In many real life situations, including job and loan applications, gatekeepers must make justified, real-time decisions about a person's fitness for a particular opportunity. People on both sides of such decisions have understandable concerns about their fairness, especially when they occur online or algorithmically. In this paper we consider the setting where we try to satisfy approximate fairness in an online decision making process where examples are sampled i.i.d from an underlying distribution. The fairness metric we consider is "equalized odds", which requires that approximately equalized false positive rates and false negative rates across groups. Our work follows from the classical learning from experts scheme and extends the multiplicative weights algorithm by maintaining an estimation for label distribution and keeping separate weights for label classes as well as groups. Our theoretical results show that approximate equalized odds can be achieved without sacrificing much regret from some distributions. We also demonstrate the algorithm on real data sets commonly used by the fairness community.

--Clustering is an effective technique in data mining to group a set of objects in terms of some attributes. Theoretical analyses and extensive experiments on several public datasets demonstrate the effectiveness and rationality of our proposed REFCMFS method. S a fundamental problem in machine learning, clustering is widely used for many fields, such as the network data (including Protein-Protein Interaction Networks [1], Road Networks [2], Geo-Social Network [3]), medical diagnosis [4], biological data analysis [5], environmental chemistry [6] and so on. K-Means clustering is one of the most popular techniques because of its simplicity and effectiveness, which randomly initializes the cluster centroids, assigns each sample to its nearest cluster and then updates cluster centroid itera-tively to cluster a dataset into some subsets. Over the past years, many modified versions of K-Means algorithms have been proposed, such as K-Means based Consensus clustering [7], Optimized Cartesian K-Means [8], Group K-Means [9] and so on. Jinglin Xu and Junwei Han were with the School of Automation, Northwestern Polytechnical University, Xi'an 710072, Shaanxi, China. Feiping Nie is with School of Computer Science and Center for OPTical IMagery Analysis and Learning (OPTIMAL), Northwestern Polytechnical University, Xi'an 710072, Shaanxi, China. Xuelong Li is with School of Computer Science and Center for OPTical IMagery Analysis and Learning (OPTIMAL), Northwestern Polytechnical University, Xi'an 710072, Shaanxi, China.

A massive number of well-trained deep networks have been released by developers online. These networks may focus on different tasks and in many cases are optimized for different datasets. In this paper, we study how to exploit such heterogeneous pre-trained networks, known as teachers, so as to train a customized student network that tackles a set of selective tasks defined by the user . W e assume no human annotations are available, and each teacher may be either single-or multi-task. T o this end, we introduce a dual-step strategy that first extracts the task-specific knowledge from the heterogeneous teachers sharing the same sub-task, and then amalgamates the extracted knowledge to build the student network. T o facilitate the training, we employ a selective learning scheme where, for each unlabelled sample, the student learns adaptively from only the teacher with the least prediction ambiguity. W e evaluate the proposed approach on several datasets and experimental results demonstrate that the student, learned by such adaptive knowledge amalgamation, achieves performances even better than those of the teachers.

Successful robot-assisted feeding requires bite acquisition of a wide variety of food items. Different food items may require different manipulation actions for successful bite acquisition. Therefore, a key challenge is to handle previously-unseen food items with very different action distributions. By leveraging contexts from previous bite acquisition attempts, a robot should be able to learn online how to acquire those previously-unseen food items. In this ongoing work, we construct a contextual bandit framework for this problem setting. We then propose variants of the $\epsilon$-greedy and LinUCB contextual bandit algorithms to minimize cumulative regret within that setting. In future, we expect empirical estimates of cumulative regret for each algorithm on robot bite acquisition trials as well as updated theoretical regret bounds that leverage the more structured context of this problem setting.

Despite numerous research work in reinforcement learning (RL) and the recent successes obtained by combining it with deep learning, deep reinforcement learning (DRL) is still facing many challenges. Some of them, like the ability to abstract actions or the difficulty to explore the environment with sparse rewards, can be addressed by the use of intrinsic motivation. In this article, we provide a survey on the role of intrinsic motivation in DRL. We categorize the different kinds of intrinsic motivations and detail their interests and limitations. Our investigation shows that the combination of DRL and intrinsic motivation enables to learn more complicated and more generalisable behaviours than standard DRL. We provide an in-depth analysis describing learning modules through an unifying scheme composed of information theory, compression theory and reinforcement learning. We then explain how these modules could serve as building blocks over a complete developmental architecture, highlighting the numerous outlooks of the domain.

This papers aims to examine the potential of using the emerging deep reinforcement learning techniques in flight control. Instead of learning from scratch, the autopilot structure is fixed as typical three-loop autopilot and deep reinforcement learning is utilised to learn the autopilot gains. This domain-knowledge-aided approach is proved to significantly improve the learning efficiency. To solve the flight control problem, we then formulate a Markovian decision process with a proper reward function that enable the application of reinforcement learning theory. The state-of-the-art deep deterministic policy gradient algorithm is utilised to learn an action policy that maps the observed states to the autopilot gains. Extensive empirical numerical simulations are performed to validate the proposed computational control algorithm.

Do I need to attend the live lectures? All lectures are recorded and shared with the class, so you can catch up whenever your schedule allows. Will there be homework or quizzes?

Mathematically, this is why we need to understand partial derivatives, since they allow us to compute the relationship between components of the neural network and the cost function. And as should be obvious, we want to minimize the cost function. When we know what affects it, we can effectively change the relevant weights and biases to minimize the cost function. If you are not a math student or have not studied calculus, this is not at all clear. So let me try to make it more clear. The squished'd' is the partial derivative sign.