The educational system is broken, and unfair. For decades, if not centuries, learning was limited by geography and having the means to continue with higher education. Online learning and massive open online courses (MOOCs) promised to address the inequities in education while extending its reach across all geographies. However, the online model simply paved over the older methods with technology, and perhaps even making things worse -- pushing course material to students, with no effective way to track how much they're learning, or even if they're paying attention. Now, artificial intelligence (AI) may have an answer for that, bringing learning and feedback in a very personal way to students.

Current knowledge distillation methods require full training data to distill knowledge from a large "teacher" network to a compact "student" network by matching certain statistics between "teacher" and "student" such as softmax outputs and feature responses. This is not only time-consuming but also inconsistent with human cognition in which children can learn knowledge from adults with few examples. This paper proposes a novel and simple method for knowledge distillation from few samples. Taking the assumption that both "teacher" and "student" have the same feature map sizes at each corresponding block, we add a 1x1 conv-layer at the end of each block in the student-net, and align the block-level outputs between "teacher" and "student" by estimating the parameters of the added layer with limited samples. We prove that the added layer can be absorbed/merged into the previous conv-layer to formulate a new conv-layer with the same size of parameters and computation cost as the previous one. Experiments verify that the proposed method is very efficient and effective to distill knowledge from teacher-net to student-net constructing in different ways on various datasets.

The capacity of meta-learning algorithms to quickly adapt to a variety of tasks, including ones they did not experience during meta-training, has been a key factor in the recent success of these methods on few-shot learning problems. This particular advantage of using meta-learning over standard supervised or reinforcement learning is only well founded under the assumption that the adaptation phase does improve the performance of our model on the task of interest. However, in the classical framework of meta-learning, this constraint is only mildly enforced, if not at all, and we only see an improvement on average over a distribution of tasks. In this paper, we show that the adaptation in an algorithm like MAML can significantly decrease the performance of an agent in a meta-reinforcement learning setting, even on a range of meta-training tasks.

While machine learning is traditionally a resource intensive task, embedded systems, autonomous navigation and the vision of the Internet-of-Things fuel the interest in resource efficient approaches. These approaches require a carefully chosen trade-off between performance and resource consumption in terms of computation and energy. On top of this, it is crucial to treat uncertainty in a consistent manner in all but the simplest applications of machine learning systems. In particular, a desideratum for any real-world system is to be robust in the presence of outliers and corrupted data, as well as being `aware' of its limits, i.e.\ the system should maintain and provide an uncertainty estimate over its own predictions. These complex demands are among the major challenges in current machine learning research and key to ensure a smooth transition of machine learning technology into every day's applications. In this article, we provide an overview of the current state of the art of machine learning techniques facilitating these real-world requirements. First we provide a comprehensive review of resource-efficiency in deep neural networks with focus on techniques for model size reduction, compression and reduced precision. These techniques can be applied during training or as post-processing and are widely used to reduce both computational complexity and memory footprint. As most (practical) neural networks are limited in their ways to treat uncertainty, we contrast them with probabilistic graphical models, which readily serve these desiderata by means of probabilistic inference. In that way, we provide an extensive overview of the current state-of-the-art of robust and efficient machine learning for real-world systems.

Modeling a sequence of interactions between users and items (e.g., products, posts, or courses) is crucial in domains such as e-commerce, social networking, and education to predict future interactions. Representation learning presents an attractive solution to model the dynamic evolution of user and item properties, where each user/item can be embedded in a euclidean space and its evolution can be modeled by dynamic changes in embedding. However, existing embedding methods either generate static embeddings, treat users and items independently, or are not scalable. Here we present JODIE, a coupled recurrent model to jointly learn the dynamic embeddings of users and items from a sequence of user-item interactions. JODIE has three components. First, the update component updates the user and item embedding from each interaction using their previous embeddings with the two mutually-recursive Recurrent Neural Networks. Second, a novel projection component is trained to forecast the embedding of users at any future time. Finally, the prediction component directly predicts the embedding of the item in a future interaction. For models that learn from a sequence of interactions, traditional training data batching cannot be done due to complex user-user dependencies. Therefore, we present a novel batching algorithm called t-Batch that generates time-consistent batches of training data that can run in parallel, giving massive speed-up. We conduct six experiments on two prediction tasks---future interaction prediction and state change prediction---using four real-world datasets. We show that JODIE outperforms six state-of-the-art algorithms in these tasks by up to 22.4%. Moreover, we show that JODIE is highly scalable and up to 9.2x faster than comparable models. As an additional experiment, we illustrate that JODIE can predict student drop-out from courses five interactions in advance.

More accurate machine learning models often demand more computation and memory at test time, making them difficult to deploy on CPU- or memory-constrained devices. Model compression (also known as distillation) alleviates this burden by training a less expensive student model to mimic the expensive teacher model while maintaining most of the original accuracy. However, when fresh data is unavailable for the compression task, the teacher's training data is typically reused, leading to suboptimal compression. In this work, we propose to augment the compression dataset with synthetic data from a generative adversarial network (GAN) designed to approximate the training data distribution. Our GAN-assisted model compression (GAN-MC) significantly improves student accuracy for expensive models such as deep neural networks and large random forests on both image and tabular datasets. Building on these results, we propose a comprehensive metric---the Compression Score---to evaluate the quality of synthetic datasets based on their induced model compression performance. The Compression Score captures both data diversity and discriminability, and we illustrate its benefits over the popular Inception Score in the context of image classification.

Ricardo Diaz is a machine learning engineer. He works for a great company in Peru, and he's a graduate of no less than four Nanodegree programs! But just two years ago, it was a different story. He was still in Venezuela, struggling to learn new skills. He was short of money, and his prospects for making a full-time salary weren't great.

In January 2014, Stanford University professors Trevor Hastie and Rob Tibshirani (authors of the legendary Elements of Statistical Learning textbook) taught an online course based on their newest textbook, An Introduction to Statistical Learning with Applications in R (ISLR). I found it to be an excellent course in statistical learning (also known as "machine learning"), largely due to the high quality of both the textbook and the video lectures. And as an R user, it was extremely helpful that they included R code to demonstrate most of the techniques described in the book. If you are new to machine learning (and even if you are not an R user), I highly recommend reading ISLR from cover-to-cover to gain both a theoretical and practical understanding of many important methods for regression and classification. It is available as a free PDF download from the authors' website.

According to the World Economics Forum's The Future of Jobs 2018 report, machines will overtake humans in terms of performing more tasks at the workplace by 2025 -- but there could still be 58 million net new jobs created in the next five years. The report notes that the growing skills for 2022 will include analytical thinking, creativity, critical thinking, complex problem solving, and systems analysis. Also: Can humans get a handle on AI? The Future of Jobs Report 2018 also identified 10 emerging jobs in 2022, including data analysts and scientists, AI and machine learning specialists and general and operation managers as the top 3 jobs. AI and advancements in automation may result in 75 million job displacements, but at the same time period another 133 million new roles will emerge where people and machines will co-exist, creating a net new 58 million jobs by 2022.

As researchers and practitioners of applied machine learning, we are given a set of requirements on the problem to be solved, the plausibly obtainable data, and the computational resources available. We aim to find (within those bounds) reliably useful combinations of problem, data, and algorithm. An emphasis on algorithmic or technical novelty in ML conference publications leads to exploration of one dimension of this space. Data collection and ML deployment at scale in industry settings offers an environment for exploring the others. Our conferences and reviewing criteria can better support empirical ML by soliciting and incentivizing experimentation and synthesis independent of algorithmic innovation.