The greater degree of control and personalization offered by learning apps and online platforms promise to facilitate the design and implementation of automated, data-driven teaching policies that adapt to each learner's knowledge over time, improving upon the traditional one-size-fits-all human instruction. However, to fulfill this promise, it is necessary to develop adaptive data-driven models of the learners, which accurately quantify their knowledge, and efficient methods to find teaching policies that are provably optimal under the learners' models [1, 2]. In this context, research in the (theoretical) computer science literature has been typically focused on finding teaching policies that enjoy optimality guarantees under simplified mathematical models of the learner's knowledge [3-7]. In contrast, research in cognitive sciences has focused on measuring the effectivity of a variety of heuristic teaching policies informed by psychologically valid models of the learner's knowledge using (small) randomized control trials [8-11]. Only very recently, Tabibian et al. [12] has introduced a machine learning modeling framework that bridges the gap between both lines of research--their framework can be used to determine the optimal rate of study a learner should follow under a model of the learner's memory state that is informed by real human memory data. However, in the evaluation of their framework, the authors resort to a natural experiment using data from a popular language-learning online platform rather than a randomized control trial, the gold standard in the cognitive sciences literature. As a result, it has been argued that, in an interventional setting, an actual learner following the rate of study may fail to achieve optimal performance [1]. In this paper, we build upon the modeling framework of Tabibian et al. [12] and design Select, a simple, efficient and adaptive machine learning algorithm with theoretical guarantees to determine which questions to include in a learner's sessions of study over time, rather than optimizing the rate of study as in Tabibian et al.,

Hierarchical Multi-Agent Systems provide a convenient and relevant way to analyze, model, and simulate complex systems in which a large number of entities are interacting at different levels of abstraction. In this paper, we introduce HAMLET (Hierarchical Agent-based Machine LEarning plaTform), a platform based on hierarchical multi-agent systems, to facilitate the research and democratization of machine learning entities distributed geographically or locally. This is carried out by firstly modeling the machine learning solutions as a hypergraph and then autonomously setting up a multi-level structure composed of heterogeneous agents based on their innate capabilities and learned skills. HAMLET aids the design and management of machine learning systems and provides analytical capabilities for the research communities to assess the existing and/or new algorithms/datasets through flexible and customizable queries. The proposed platform does not assume restrictions on the type of machine learning algorithms/datasets and is theoretically proven to be sound and complete with polynomial computational requirements. Additionally, it is examined empirically on 120 training and four generalized batch testing tasks performed on 24 machine learning algorithms and 9 standard datasets. The experimental results provided not only establish confidence in the platform's consistency and correctness but also demonstrates its testing and analytical capacity.

Continual (sequential) training and multitask (simultaneous) training are often attempting to solve the same overall objective: to find a solution that performs well on all considered tasks. The main difference is in the training regimes, where continual learning can only have access to one task at a time, which for neural networks typically leads to catastrophic forgetting. That is, the solution found for a subsequent task does not perform well on the previous ones anymore. However, the relationship between the different minima that the two training regimes arrive at is not well understood. Is there a local structure that could explain the difference in performance achieved by the two different schemes? Motivated by recent work showing that different minima of the same task are typically connected by very simple curves of low error, we investigate whether multitask and continual solutions are similarly connected. We empirically find that indeed such connectivity can be reliably achieved and, more interestingly, it can be done by a linear path, conditioned on having the same initialization for both. We thoroughly analyze this observation and discuss its significance for the continual learning process. Furthermore, we exploit this finding to propose an effective algorithm that constrains the sequentially learned minima to behave as the multitask solution. One major consequence of learning multiple tasks in a continual learning (CL) setting -- where tasks are learned sequentially, and the model can only have access to one task at a time -- is catastrophic forgetting (McCloskey & Cohen, 1989).

The goal of text generation is to make machines express in human language. It is one of the most important yet challenging tasks in natural language processing (NLP). Since 2014, various neural encoder-decoder models pioneered by Seq2Seq have been proposed to achieve the goal by learning to map input text to output text. However, the input text alone often provides limited knowledge to generate the desired output, so the performance of text generation is still far from satisfaction in many real-world scenarios. To address this issue, researchers have considered incorporating various forms of knowledge beyond the input text into the generation models. This research direction is known as knowledge-enhanced text generation. In this survey, we present a comprehensive review of the research on knowledge enhanced text generation over the past five years. The main content includes two parts: (i) general methods and architectures for integrating knowledge into text generation; (ii) specific techniques and applications according to different forms of knowledge data. This survey can have broad audiences, researchers and practitioners, in academia and industry.

With wit, vision, and scholarship, Domingos describes how these scientists are creating programs that allow a computer to teach itself.

Editor's note: Astrobites is a graduate-student-run organization that digests astrophysical literature for undergraduate students. We hope you enjoy this post from astrobites; the original can be viewed at astrobites.org. Galaxy clusters are among the largest gravitationally bound structures in the universe. One of their defining characteristics is that they tend to be embedded within a large reservoir of superheated gas, known as the intracluster medium (ICM). With temperatures up to 108 Kelvin, the ICM is a strong emitter of X-ray radiation.

The A-level and GCSE exams fiasco exposes the significant problems in using algorithms to make complicated, life-changing decisions. As we rely more on AI and automation, "smart" systems are as likely to create as many problems as they solve. AI is not all bad nor will recent problems pause its growth, rather organisations continue to hit the accelerator on AI technologies on the promise of rapid delivery of value. So, does a fast percolation of AI technologies throughout business and public sector bodies mean we can dispose of the human talent in their analytics teams? No. Experience shows people can comfortably integrate new technologies into their jobs and lives; AI and automation are no exception.

Inaugurating a five-day global virtual summit on artificial intelligence (AI), Responsible AI for Social Empowerment or RAISE 2020, Prime Minister Narendra Modi highlighted how the use of AI will empower India and also warned against the pitfalls. In June, India along with Australia, the United States, the United Kingdom, Canada, France, Germany, New Zealand and others joined together to create the Global Partnership on Artificial Intelligence (GPAI) for the responsible development and use of AI. Algorithm transparency is key to establishing this trust. This will create an e-education unit to boost the digital infrastructure, digital content and capacity. Under this programme, more than 11,000 students from schools completed the basic course.

We have recently shared introductory courses from Amazon's recently-launched Machine Learning University initiative, courses focusing on both computer vision (CV) and natural language processing (NLP). These courses assume no previous knowledge of the topics, and are based on short video lectures and corresponding Python notebooks. But what if you are not interested in the specialized paths of either CV or NLP? What if you would like to gain an understanding of applying machine learning to more traditional datasets? If this remark resembles you, Accelerated Tabular Data from Amazon's Machine Learning University might be a good place to start.

Trainers and coaches can use Exer Studio to provide more interaction and direct feedback in virtual classes. The pandemic happened to be perfect timing for the launch of three friends' artificial intelligence fitness platform. Denver-based Exer Labs, which launched its first app in May, uses AI on mobile and tablet devices as well as computer vision to help fitness users experience a more interactive and affordable at-home workout. "Our goal is to not make fitness such a luxury," said Exer co-founder Zaw Thet. "We want to focus on the hardware that people already have at home and make it accessible for everyone."