Despite years of hype (and plenty of worries) about the all-conquering power of Artificial Intelligence (AI), there still remains a significant gap between the promise of AI and its reality for business. Tech firms have pitched AI's capabilities for years, but for most organisations, the benefits of AI remain elusive. It's hard to gauge the proportion of businesses that are effectively using artificial intelligence today, and to what extent. Adoption rates shown in recent reports fall anywhere between 20% and 30%, with adoption typically loosely defined as "implementing AI in some form". A survey led by KPMG among 30 of the Global 500 companies found that although 30% of respondents reported using AI for a selective range of functions, only 17% of the companies were deploying the technology "at scale" within the enterprise.

For all its successes, Reinforcement Learning (RL) still struggles to deliver formal guarantees on the closed-loop behavior of the learned policy. Among other things, guaranteeing the safety of RL with respect to safety-critical systems is a very active research topic. Some recent contributions propose to rely on projections of the inputs delivered by the learned policy into a safe set, ensuring that the system safety is never jeopardized. Unfortunately, it is unclear whether this operation can be performed without disrupting the learning process. This paper addresses this issue. The problem is analysed in the context of $Q$-learning and policy gradient techniques. We show that the projection approach is generally disruptive in the context of $Q$-learning though a simple alternative solves the issue, while simple corrections can be used in the context of policy gradient methods in order to ensure that the policy gradients are unbiased. The proposed results extend to safe projections based on robust MPC techniques.

This course is designed to equip you with the theoretical and practical knowledge of Machine Learning as applied for geospatial analysis, namely Geographic Information Systems (GIS) and Remote Sensing. By the end of the course, you will feel confident and completely understand the Machine Learning applications in GIS technology and how to use Machine Learning algorithms for various geospatial tasks, such as land use and land cover mapping (classifications) and object-based image analysis (segmentation). This course will also prepare you for using GIS with open source and free software tools. In the course, you will be able to apply such Machine Learning algorithms as Random Forest, Support Vector Machines and Decision Trees (and others) for classification of satellite imagery. On top of that, you will practice GIS by completing an entire GIS project by exploring the power of Machine Learning, cloud computing and Big Data analysis using Google Erath Engine for any geographic area in the world.

Scene classification is an important aspect of image/video understanding and segmentation. However, remote-sensing scene classification is a challenging image recognition task, partly due to the limited training data, which causes deep-learning Convolutional Neural Networks (CNNs) to overfit. Another difficulty is that images often have very different scales and orientation (viewing angle). Yet another is that the resulting networks may be very large, again making them prone to overfitting and unsuitable for deployment on memory- and energy-limited devices. We propose an efficient deep-learning approach to tackle these problems. We use transfer learning to compensate for the lack of data, and data augmentation to tackle varying scale and orientation.

Public perceptions of Robotics and Artificial Intelligence (RAI) are important in the acceptance, uptake, government regulation and research funding of this technology. Recent research has shown that the public's understanding of RAI can be negative or inaccurate. We believe effective public engagement can help ensure that public opinion is better informed. In this paper, we describe our first iteration of a high throughput in-person public engagement activity. We describe the use of a light touch quiz-format survey instrument to integrate in-the-wild research participation into the engagement, allowing us to probe both the effectiveness of our engagement strategy, and public perceptions of the future roles of robots and humans working in dangerous settings, such as in the off-shore energy sector. We critique our methods and share interesting results into generational differences within the public's view of the future of Robotics and AI in hazardous environments. These findings include that older peoples' views about the future of robots in hazardous environments were not swayed by exposure to our exhibit, while the views of younger people were affected by our exhibit, leading us to consider carefully in future how to more effectively engage with and inform older people.

Real-world problems often involve the optimization of several objectives under multiple constraints. Furthermore, we may not have an expression for each objective or constraint; they may be expensive to evaluate; and the evaluations can be noisy. These functions are referred to as black-boxes. Bayesian optimization (BO) can efficiently solve the problems described. For this, BO iteratively fits a model to the observations of each black-box. The models are then used to choose where to evaluate the black-boxes next, with the goal of solving the optimization problem in a few iterations. In particular, they guide the search for the problem solution, and avoid evaluations in regions of little expected utility. A limitation, however, is that current BO methods for these problems choose a point at a time at which to evaluate the black-boxes. If the expensive evaluations can be carried out in parallel (as when a cluster of computers is available), this results in a waste of resources. Here, we introduce PPESMOC, Parallel Predictive Entropy Search for Multi-objective Optimization with Constraints, a BO strategy for solving the problems described. PPESMOC selects, at each iteration, a batch of input locations at which to evaluate the black-boxes, in parallel, to maximally reduce the entropy of the problem solution. To our knowledge, this is the first batch method for constrained multi-objective BO. We present empirical evidence in the form of synthetic, benchmark and real-world experiments that illustrate the effectiveness of PPESMOC.

We propose spectral methods for long-term forecasting of temporal signals stemming from linear and nonlinear quasi-periodic dynamical systems. For linear signals, we introduce an algorithm with similarities to the Fourier transform but which does not rely on periodicity assumptions, allowing for forecasting given potentially arbitrary sampling intervals. We then extend this algorithm to handle nonlinearities by leveraging Koopman theory. The resulting algorithm performs a spectral decomposition in a nonlinear, data-dependent basis. The optimization objective for both algorithms is highly non-convex. However, expressing the objective in the frequency domain allows us to compute global optima of the error surface in a scalable and efficient manner, partially by exploiting the computational properties of the Fast Fourier Transform. Because of their close relation to Bayesian Spectral Analysis, uncertainty quantification metrics are a natural byproduct of the spectral forecasting methods.

In this paper, we consider the problem of certifying the robustness of neural networks to perturbed and adversarial input data. Such certification is imperative for the application of neural networks in safety-critical decision-making and control systems. Certification techniques using convex optimization have been proposed, but they often suffer from relaxation errors that void the certificate. Our work exploits the structure of ReLU networks to improve relaxation errors through a novel partition-based certification procedure. The proposed method is proven to tighten existing linear programming relaxations, and asymptotically achieves zero relaxation error as the partition is made finer. We develop a finite partition that attains zero relaxation error and use the result to derive a tractable partitioning scheme that minimizes the worst-case relaxation error. Experiments using real data show that the partitioning procedure is able to issue robustness certificates in cases where prior methods fail. Consequently, partition-based certification procedures are found to provide an intuitive, effective, and theoretically justified method for tightening existing convex relaxation techniques.

Machine learning models are omnipresent for predictions on big data. One challenge of deployed models is the change of the data over time--a phenomenon called concept drift. If not handled correctly, a concept drift can lead to significant mispredictions. We explore a novel approach for concept drift handling, which depicts a strategy to switch between the application of simple and complex machine learning models for regression tasks. We assume that the approach plays out the individual strengths of each model, switching to the simpler model if a drift occurs and switching back to the complex model for typical situations. We instantiate the approach on a real-world data set of taxi demand in New York City, which is prone to multiple drifts, e.g. the weather phenomena of blizzards, resulting in a sudden decrease of taxi demand. We are able to show that our suggested approach outperforms all regarded baselines significantly.