Constraint-based learning reduces the burden of collecting labels by having users specify general properties of structured outputs, such as constraints imposed by physical laws. We propose a novel framework for simultaneously learning these constraints and using them for supervision, bypassing the difficulty of using domain expertise to manually specify constraints. Learning requires a black-box simulator of structured outputs, which generates valid labels, but need not model their corresponding inputs or the input-label relationship. At training time, we constrain the model to produce outputs that cannot be distinguished from simulated labels by adversarial training. Providing our framework with a small number of labeled inputs gives rise to a new semi-supervised structured prediction model; we evaluate this model on multiple tasks --- tracking, pose estimation and time series prediction --- and find that it achieves high accuracy with only a small number of labeled inputs. In some cases, no labels are required at all.

Deep learning is the mainstream technique for many machine learning tasks, including image recognition, machine translation, speech recognition, and so on. It has outperformed conventional methods in various fields and achieved great successes. Unfortunately, the understanding on how it works remains unclear. It has the central importance to lay down the theoretic foundation for deep learning. In this work, we give a geometric view to understand deep learning: we show that the fundamental principle attributing to the success is the manifold structure in data, namely natural high dimensional data concentrates close to a low-dimensional manifold, deep learning learns the manifold and the probability distribution on it. We further introduce the concepts of rectified linear complexity for deep neural network measuring its learning capability, rectified linear complexity of an embedding manifold describing the difficulty to be learned. Then we show for any deep neural network with fixed architecture, there exists a manifold that cannot be learned by the network. Finally, we propose to apply optimal mass transportation theory to control the probability distribution in the latent space.

Future 5G wireless networks will rely on agile and automated network management, where the usage of diverse resources must be jointly optimized with surgical accuracy. A number of key wireless network functionalities (e.g., traffic steering, energy savings) give rise to hard optimization problems. What is more, high spatio-temporal traffic variability coupled with the need to satisfy strict per slice/service SLAs in modern networks, suggest that these problems must be constantly (re-)solved, to maintain close-to-optimal performance. To this end, in this paper we propose the framework of Online Network Optimization (ONO), which seeks to maintain both agile and efficient control over time, using an arsenal of data-driven, adaptive, and AI-based techniques. Since the mathematical tools and the studied regimes vary widely among these methodologies, a theoretical comparison is often out of reach. Therefore, the important question "what is the right ONO technique?" remains open to date. In this paper, we discuss the pros and cons of each technique and further attempt a direct quantitative comparison for a specific use case, using real data. Our results suggest that carefully combining the insights of problem modeling with state-of-the-art AI techniques provides significant advantages at reasonable complexity.

Human lives are important. The decision to allow self-driving vehicles operate on our roads carries great weight. This has been a hot topic of debate between policy-makers, technologists and public safety institutions. The recent Uber Inc. self-driving car crash, resulting in the death of a pedestrian, has strengthened the argument that autonomous vehicle technology is still not ready for deployment on public roads. In this work, we analyze the Uber car crash and shed light on the question, "Could the Uber Car Crash have been avoided?". We apply state-of-the-art Computer Vision models to this highly practical scenario. More generally, our experimental results are an evaluation of various image enhancement and object recognition techniques for enabling pedestrian safety in low-lighting conditions using the Uber crash as a case study.

Needless to say, mathematics has become the reliable foundation of modern natural science, including several branches of theoretical computer science, by justifying theorems with proofs. The importance of correct proofs leads to the study of software called proof assistants [Nipkow et al. 2002; Norell 2009; The Coq Development Team 2017], which allow users to state theorems and their proofs formally in the form of certain programming languages and automatically check that the proofs correctly prove the theorems. The realm of the areas that rely on theorem proving is expanding beyond mathematics; for example, it is being applied for system verification [Klein et al. 2009; Leroy 2009], where one states the correctness of a system as a theorem and justifies it in the form of proofs. Automated theorem proving (ATP) [Bibel 2013; Fitting 2012; Pfenning 2004] is a set of techniques that prove logical formulas automatically. We are concerned with the following form of ATP called automated proof synthesis (APS): Given a logical formula P, if P holds, return a proof M of P. In the light of the importance of theorem proving, APS serves as a useful tool for activities based on formal reasoning. For example, from the perspective of the aforementioned system verification, APS serves for automating system verification; indeed, various methods for (semi)automated static program verification [Barnett et al. 2005; Chalin et al. 2007; Filliâtre and Paskevich 2013] can be seen as APS procedures. We also remark another important application of APS: automated program synthesis. An APS algorithm can be seen as an automated program synthesis procedure via the Curry-Howard isomorphism [Sørensen and Urzyczyn 2006], in which M can be seen as a program and P can be seen as a specification. Not only is APS interesting from the practical viewpoint, it is also interesting from the theoretical perspective in that it investigates the algorithmic aspect of theorem proving.

Many works in collaborative robotics and human-robot interaction focuses on identifying and predicting human behaviour while considering the information about the robot itself as given. This can be the case when sensors and the robot are calibrated in relation to each other and often the reconfiguration of the system is not possible, or extra manual work is required. We present a deep learning based approach to remove the constraint of having the need for the robot and the vision sensor to be fixed and calibrated in relation to each other. The system learns the visual cues of the robot body and is able to localise it, as well as estimate the position of robot joints in 3D space by just using a 2D color image. The method uses a cascaded convolutional neural network, and we present the structure of the network, describe our own collected dataset, explain the network training and achieved results. A fully trained system shows promising results in providing an accurate mask of where the robot is located and a good estimate of its joints positions in 3D. The accuracy is not good enough for visual servoing applications yet, however, it can be sufficient for general safety and some collaborative tasks not requiring very high precision. The main benefit of our method is the possibility of the vision sensor to move freely. This allows it to be mounted on moving objects, for example, a body of the person or a mobile robot working in the same environment as the robots are operating in.

This methodology is probably the most appropriate and the different phases provide a strong framework to practitioners but, in order to support more specific explanation of how it apply to "classic" Machine Learning and Deep Learning I needed to complement CRISP-DM with more specific design flow: Since the aim of the flows is to explain the difference between the three approaches, some (lot of…) are not incorporated to the flows, but I think it would help those who intend to introduce Machine Learning and Deep Learning to not specialists. Bio: Stéphane Faure is an IT professional at IBM where he supports server sales across Europe. In the last years, he has been working on payment fraud detections, credit scoring and regularly presents and teaches predictive analytics.

This year, the DJ, artist, and Qosmo CEO Nao Tokui flipped the concept on its head. His project, Imaginary Soundscapes, is a convolutional neural network that hears sounds when it looks at images. Based on a given image, the software will choose from 15,000 sound files to find the "soundscape" that fits. First, he applied the software to Google StreetView to create an audio tour of the world, with AI-generated sounds to accompany any scene from StreetView, from echoing voices in Barcelona's Sagrada Familia cathedral to chirping birds on rural backroads. Viewers could "immerse themselves into the artificial soundscape'imagined' by our deep learning models," Tokui explained on Medium.

The endoscope and colonoscope were first developed in 1880s to look inside the body. Specialists use their expertise and experience to examine the medical images. But sometimes, human error and backend issues can result in misdiagnosis. Population increase and more cases of internal diseases are overloading the medical industry in many major cities in the world. In turn, the demand of medical specialists continues to soar.

A computer was better than human dermatologists at detecting skin cancer in a study that pitted people against machines in the quest for better, faster diagnostics, researchers said on Tuesday. A team from Germany, the United States and France taught an artificial intelligence system to distinguish dangerous skin lesions from benign ones, showing it more than 100,000 images. The machine – a deep learning convolutional neural network or CNN – was then tested against 58 dermatologists from 17 countries, shown photos of malignant melanomas and benign moles. Just over half the dermatologists were at "expert" level with more than five years of experience, 19% had between two and five years' experience, and 29% were beginners with less than two years under their belt. "Most dermatologists were outperformed by the CNN," the research team wrote in a paper published in the journal Annals of Oncology.