A BSTRACT Current deep neural networks can achieve remarkable performance on a single task. However, when the deep neural network is continually trained on a sequence of tasks, it seems to gradually forget the previous learned knowledge. This phenomenon is referred to as catastrophic forgetting and motivates the field called lifelong learning. The central question in lifelong learning is how to enable deep neural networks to maintain performance on old tasks while learning a new task. In this paper, we introduce a novel and effective lifelong learning algorithm, called MixEd stochastic GrAdient (MEGA), which allows deep neural networks to acquire the ability of retaining performance on old tasks while learning new tasks. Extensive experimental results show that the proposed MEGA algorithm significantly advances the state-of-the-art on all four commonly used lifelong learning benchmarks, reducing the error by up to 18%. 1 I NTRODUCTION A significant step towards artificial general intelligence (AGI) is to enable the learning agent to acquire the ability of remembering past experiences while being trained on a continuum of tasks. Current deep neural networks are capable of achieving remarkable performance on a single task (Goodfellow et al., 2016). However when the network is retrained on a new task, its performance drops drastically on previously trained tasks, a phenomenon which is referred to as catastrophic forgetting (Ratcliff, 1990; Robins, 1995; French, 1999; Kirkpatrick et al., 2017).

Meta-learning has been widely used for implementing few-shot learning and fast model adaptation. One kind of meta-learning methods attempt to learn how to control the gradient descent process in order to make the gradient-based learning have high speed and generalization. This work proposes a method that controls the gradient descent process of the model parameters of a neural network by limiting the model parameters in a low-dimensional latent space. The main challenge of this idea is that a decoder with too many parameters is required. This work designs a decoder with typical structure and shares a part of weights in the decoder to reduce the number of the required parameters. Besides, this work has introduced ensemble learning to work with the proposed approach for improving performance. The results show that the proposed approach is witnessed by the superior performance over the Omniglot classification and the miniImageNet classification tasks.

Domain shifts are frequently encountered in real-world scenarios. In this paper, we consider the problem of domain-invariant deep learning by explicitly modeling domain shifts with only a small amount of domain-specific parameters in a Convolutional Neural Network (CNN). By exploiting the observation that a convolutional filter can be well approximated as a linear combination of a small set of basis elements, we show for the first time, both empirically and theoretically, that domain shifts can be effectively handled by decomposing a regular convolutional layer into a domain-specific basis layer and a domain-shared basis coefficient layer, while both remain convolutional. An input channel will now first convolve spatially only with each respective domain-specific basis to "absorb" domain variations, and then output channels are linearly combined using common basis coefficients trained to promote shared semantics across domains. We use toy examples, rigorous analysis, and real-world examples to show the framework's effectiveness in cross-domain performance and domain adaptation. With the proposed architecture, we need only a small set of basis elements to model each additional domain, which brings a negligible amount of additional parameters, typically a few hundred. 1 Introduction Training supervised deep networks requires large amount of labeled training data; however, well-trained deep networks often degrade dramatically on testing data from a significantly different domain. In real-world scenarios, such domain shifts are introduced by many factors, such as different illumination, viewing angles, and resolutions. Research topics such as transfer learning and domain adaptation are studied to promote invariant representations across domains with different levels of availabilities of annotated data. Recent efforts on learning cross-domain invariant representations using deep networks generally fall into two categories. The first one is to learn a common network with constraints encouraging invariant feature representations across different domains [11, 14, 26].

The MIT Stephen A. Schwarzman College of Computing (SCC) will reorient the Institute to bring the power of computing and artificial intelligence to all fields at MIT, and to allow the future of computing and AI to be shaped by all MIT disciplines. To support ongoing planning for the new college, Dean Melissa Nobles invited faculty from all 14 of MIT's humanistic disciplines in the School of Humanities, Arts, and Social Sciences to respond to two questions: As Nobles says in her foreword to the series, "Together, the following responses to these two questions offer something of a guidebook to the myriad, productive ways that technical, humanistic, and scientific fields can join forces at MIT, and elsewhere, to further human and planetary well-being." The following excerpts highlight faculty responses, with links to full commentaries. The excerpts are sequenced by fields in the following order: the humanities, arts, and social sciences. "The advent of artificial intelligence presents our species with an historic opportunity -- disguised as an existential challenge: Can we stay human in the age of AI? In fact, can we grow in humanity, can we shape a more humane, more just, and sustainable world? With a sense of promise and urgency, we are embarked at MIT on an accelerated effort to more fully integrate the technical and humanistic forms of discovery in our curriculum and research, and in our habits of mind and action."

Artificial Intelligence (AI) is no sooner the next big thing, it is now a big technology in digital marketing. All digital marketing operations are now affected by AI-powered devices. From startups to large companies are opting for AI-powered digital marketing equipments to enhance campaign planning and decision making. AI-based tools are now a flourishing market, with an exceptional change demand. As per the greater part of the digital marketers AI upgrading every one of the zones where the prescient investigation, basic leadership and mechanization endeavors required.

Today, education doesn't stop after students graduate; many continue learning throughout their life. And in addition to adult learning courses at colleges and universities, a lot of courses are now offered online – ranging from MOOCs ("massive open online course") to many apps. At the Global Education & Skills Forum in March, two of the 10 finalists were lifelong learning startups. Here are four of the most promising EdTech startups around. Don't have time to read a book a day?

Voice-controlled personal and home assistants (such as the Amazon Echo and Apple Siri) are becoming increasingly popular for a variety of applications. However, the benefits of these technologies are not readily accessible to Deaf or Hard-ofHearing (DHH) users. The objective of this study is to develop and evaluate a sign recognition system using multiple modalities that can be used by DHH signers to interact with voice-controlled devices. With the advancement of depth sensors, skeletal data is used for applications like video analysis and activity recognition. Despite having similarity with the well-studied human activity recognition, the use of 3D skeleton data in sign language recognition is rare. This is because unlike activity recognition, sign language is mostly dependent on hand shape pattern. In this work, we investigate the feasibility of using skeletal and RGB video data for sign language recognition using a combination of different deep learning architectures. We validate our results on a large-scale American Sign Language (ASL) dataset of 12 users and 13107 samples across 51 signs. It is named as GMUASL51. We collected the dataset over 6 months and it will be publicly released in the hope of spurring further machine learning research towards providing improved accessibility for digital assistants.

Imitation learning seeks to learn an expert policy from sampled demonstrations. However, in the real world, it is often difficult to find a perfect expert and avoiding dangerous behaviors becomes relevant for safety reasons. We present the idea of \textit{learning to avoid}, an objective opposite to imitation learning in some sense, where an agent learns to avoid a demonstrator policy given an environment. We define avoidance learning as the process of optimizing the agent's reward while avoiding dangerous behaviors given by a demonstrator. In this work we develop a framework of avoidance learning by defining a suitable objective function for these problems which involves the \emph{distance} of state occupancy distributions of the expert and demonstrator policies. We use density estimates for state occupancy measures and use the aforementioned distance as the reward bonus for avoiding the demonstrator. We validate our theory with experiments using a wide range of partially observable environments. Experimental results show that we are able to improve sample efficiency during training compared to state of the art policy optimization and safety methods.

In this paper, we study the convergence of a class of gradient-based Model-Agnostic Meta-Learning (MAML) methods and characterize their overall computational complexity as well as their best achievable level of accuracy in terms of gradient norm for nonconvex loss functions. In particular, we start with the MAML algorithm and its first order approximation (FO-MAML) and highlight the challenges that emerge in their analysis. By overcoming these challenges not only we provide the first theoretical guarantees for MAML and FO-MAML in nonconvex settings, but also we answer some of the unanswered questions for the implementation of these algorithms including how to choose their learning rate (stepsize) and the batch size for both tasks and datasets corresponding to tasks. In particular, we show that MAML can find an $\epsilon$-first-order stationary point for any positive $\epsilon$ after at most $\mathcal{O}(1/\epsilon^2)$ iterations at the expense of requiring second-order information. We also show that the FO-MAML method which ignores the second-order information required in the update of MAML cannot achieve any small desired level of accuracy, i.e, FO-MAML cannot find an $\epsilon$-first-order stationary point for any positive $\epsilon$. We further propose a new variant of the MAML algorithm called Hessian-free MAML (HF-MAML) which preserves all theoretical guarantees of MAML, without requiring access to the second-order information of loss functions.

Immersive simulations are increasingly used for teaching and training in many societally important arenas including healthcare, disaster response and science education. The interactions of participants in such settings lead to a complex array of emergent outcomes that present challenges for analysis. This paper studies a central element of such an analysis, namely the interpretability of models for inferring structure in time series data. This problem is explored in the context of modeling student interactions in an immersive ecological-system simulation. Unsupervised machine learning is applied to data on system dynamics with the aim of helping teachers determine the effects of students' actions on these dynamics. We address the question of choosing the optimal machine learning model, considering both statistical information criteria and interpretabilty quality. The results of a user study show that the models that are the best understood by people are not those that optimize information theoretic criteria. In addition, a model using a fully Bayesian approach performed well on both statistical measures and on human-subject tests of interpretabilty, making it a good candidate for automated model selection that does not require human-in-the-loop evaluation. The results from this paper are already being used in the classroom and can inform the design of interpretable models for a broad range of socially relevant domains. 1 Introduction There is increasing evidence of the value of multi-person embodied simulations for engaging learners in a variety of applications, such as healthcare, disaster response and education (Alinier et al. 2014; Amir and Gal 2013).