With widespread applications of artificial intelligence (AI), the capabilities of the perception, understanding, decision-making and control for autonomous systems have improved significantly in the past years. When autonomous systems consider the performance of accuracy and transferability simultaneously, several AI methods, like adversarial learning, reinforcement learning (RL) and meta-learning, show their powerful performance. Here, we review the learning-based approaches in autonomous systems from the perspectives of accuracy and transferability. Accuracy means that a well-trained model shows good results during the testing phase, in which the testing set shares a same task or a data distribution with the training set. Transferability means that when an trained model is transferred to other testing domains, the accuracy is still good. Firstly, we introduce some basic concepts of transfer learning and then present some preliminaries of adversarial learning, RL and meta-learning. Secondly, we focus on reviewing the accuracy and transferability to show the advantages of adversarial learning, like generative adversarial networks (GANs), in typical computer vision tasks in autonomous systems, including image style transfer, image super-resolution, image deblurring/dehazing/rain removal, semantic segmentation, depth estimation and person re-identification. Then, we further review the performance of RL and meta-learning from the aspects of accuracy and transferability in autonomous systems, involving robot navigation and robotic manipulation. Finally, we discuss several challenges and future topics for using adversarial learning, RL and meta-learning in autonomous systems.

When you hear the words artificial intelligence, what's the first thing that comes to mind? Driverless cars, Amazon shopping, Netflix movie recommendations and trading software to help bankers. Many think of artificial intelligence in healthcare as a buzz word or just a concept that will fully develop in the near future, but has no impact in your life right now. Some other household examples of current-day technology that use AI include Siri, Alexa, Google Now – these popular speech recognition software assistants all use artificial intelligence! Recently, Alexa was cleared to handle patient information.

There is increasing awareness that the greatest problems with artificial intelligence are not primarily technical, but rather how to achieve value from the technology. This was a growing problem even in the booming economy of the last several years, but a much more important issue in the current pandemic-driven recessionary economic climate. Older AI technologies like natural language processing, and newer ones like deep learning, work well for the most part and are capable of providing considerable value to organizations that implement them. The challenges are with large-scale implementation and deployment of AI, which are necessary to achieve value. There is substantial evidence of this in surveys.

If anything can supersede the hype around artificial intelligence (AI) than it is probably Emotion AI, the irony is that the latter is the subset of AI itself. The hype around emotion AI revolves around the excitement of witnessing the mass infiltration of machines into complex world of human emotions. For too long machines have been considered as a beast that can interpret & simplify complex data but miserably falls short of replicating the same magic in the area of human emotion. However, this hypothesis and assumption is now being challenged by artificial intelligence. Emotion Ai is essentially one of emerging areas of AI where machines seek to analyze and comprehend human emotions by judging facial expressions, body language, gestures, voice tone so and so forth.

Symbolic time series representations allow for the use of algorithms from text processing and bioinformatics, which often take advantage of the discrete nature of the data. Our focus in this work is to develop a symbolic representation which is dimension reducing whilst preserving the essential shape of the time series. Our definition of shape is different from the one commonly implied in the context of time series: we focus on representing the peaks and troughs of the time series in their correct order of appearance, but we are happy to slightly stretch the time series in both the time and value directions. In other words, our focus is not necessarily on approximating the time series values at the correct time points, but on representing the local up-and-down behavior of the time series and identifying repeated motifs. This is obviously not appropriate in all applications, but we believe it is close to how humans summarize the overall behavior of a time series, and in that our representation might be useful for trend prediction, anomaly detection, and motif discovery. To illustrate, let us consider the time series shown in Figure 1. This series is sampled at equidistant time points with values t 0,t 1,...,t N R, where N 230. There are various ways of describing this time series, for example: (a) It is exactly representable as a high-dimensional vector T [t 0,t 1,...,t N ] R N 1 .

The application of "machine learning" and "artificial intelligence" has become popular within the last decade. Both terms are frequently used in science and media, sometimes interchangeably, sometimes with different meanings. In this work, we aim to clarify the relationship between these terms and, in particular, to specify the contribution of machine learning to artificial intelligence. We review relevant literature and present a conceptual framework which clarifies the role of machine learning to build (artificial) intelligent agents. Hence, we seek to provide more terminological clarity and a starting point for (interdisciplinary) discussions and future research.

Vector representations of graphs and relational structures, whether hand-crafted feature vectors or learned representations, enable us to apply standard data analysis and machine learning techniques to the structures. A wide range of methods for generating such embeddings have been studied in the machine learning and knowledge representation literature. However, vector embeddings have received relatively little attention from a theoretical point of view. Starting with a survey of embedding techniques that have been used in practice, in this paper we propose two theoretical approaches that we see as central for understanding the foundations of vector embeddings. We draw connections between the various approaches and suggest directions for future research.

Game-playing Evolutionary Algorithms, specifically Rolling Horizon Evolutionary Algorithms, have recently managed to beat the state of the art in performance across many games. However, the best results per game are highly dependent on the specific configuration of modifications and hybrids introduced over several works, each described as parameters in the algorithm. However, the search for the best parameters has been reduced to several human-picked combinations, as the possibility space has grown beyond exhaustive search. This paper presents the state of the art in Rolling Horizon Evolutionary algorithms, combining all modifications described in literature and some additional ones for a large resultant hybrid. It then uses a parameter optimiser, the N-Tuple Bandit Evolutionary Algorithm, to find the best combination of parameters in 20 games with various properties from the General Video Game AI Framework. We highlight the noisy optimisation problem resultant, as both the games and the algorithm being optimised are stochastic. We then analyse the algorithm's parameters and interesting combinations revealed through the parameter optimisation process. Lastly, we show that it is possible to automatically explore a large parameter space and find configurations which outperform the state of the art on several games.

To facilitate research in the direction of sample-efficient reinforcement learning, we held the MineRL Competition on Sample-Efficient Reinforcement Learning Using Human Priors at the Thirty-fourth Conference on Neural Information Processing Systems (NeurIPS 2019). The primary goal of this competition was to promote the development of algorithms that use human demonstrations alongside reinforcement learning to reduce the number of samples needed to solve complex, hierarchical, and sparse environments. We describe the competition and provide an overview of the top solutions, each of which uses deep reinforcement learning and/or imitation learning. We also discuss the impact of our organizational decisions on the competition as well as future directions for improvement.

FUDIPO is a project funded by the European Commission under H2020 programme, SPIRE-02-2016: "Plant-wide monitoring and control of data-intensive processes", which started on October 1st, 2016 and ends on 30th September 2020. Mälardalen University coordinates the project, and the consortium is composed of energy experts, applied mathematicians, and software engineering experts to face the SPIRE topic. The goal with FUDIPO project is to introduce AI systems into process industries. The special demands for industry are to have very robust functions and a good possibility to keep control of all functions to avoid causing new problems! This demands a structured work, but still utilising the most advanced functions to benefit from this new world, and see that European industry really stay in the forefront of production development.