Heuristics and Diagnostics for Complex Systems

While logistic sigmoid neurons are more biologically plausible than hyperbolic tangent neurons, the latter work better for training multi-layer neural networks. This paper shows that rectifying neurons are an even better model of biological neurons and yield equal or better performance than hyperbolic tangent networks in spite of the hard non-linearity and non-differentiability at zero, creating sparse representations with true zeros, which seem remarkably suitable for naturally sparse data. Even though they can take advantage of semi-supervised setups with extra-unlabeled data, deep rectifier networks can reach their best performance without requiring any unsupervised pre-training on purely supervised tasks with large labeled datasets. Hence, these results can be seen as a new milestone in the attempts at understanding the difficulty in training deep but purely supervised neural networks, and closing the performance gap between neural networks learnt with and without unsupervised pre-training. 1

We consider the problem of embedding entities and relationships of multi-relational data in low-dimensional vector spaces. Our objective is to propose a canonical model which is easy to train, contains a reduced number of parameters and can scale up to very large databases. Hence, we propose TransE, a method which models relationships by interpreting them as translations operating on the low-dimensional embeddings of the entities. Despite its simplicity, this assump-tion proves to be powerful since extensive experiments show that TransE signif-icantly outperforms state-of-the-art methods in link prediction on two knowledge bases. Besides, it can be successfully trained on a large scale data set with 1M entities, 25k relationships and more than 17M training samples. 1

The exponential increase in the availability of online reviews and recommendations makes sentiment classification an interesting topic in academic and industrial research. Reviews can span so many different domains that it is difficult to gather annotated training data for all of them. Hence, this paper studies the problem of domain adaptation for sentiment classifiers, hereby a system is trained on labeled reviews from one source domain but is meant to be deployed on another. We propose a deep learning approach which learns to extract a meaningful representation for each review in an unsupervised fashion. Sentiment classifiers trained with this high-level feature representation clearly outperform state-of-the-art methods on a benchmark composed of reviews of 4 types of Amazon products. Furthermore, this method scales well and allowed us to successfully perform domain adaptation on a larger industrial-strength dataset of 22 domains. 1.

In the context of cancer diagnosis and treatment, we consider the problem of constructing an accurate prediction rule on the basis of a relatively small number of tumor tissue samples of known type containing the expression data on very many (possibly thousands) genes. Recently, results have been presented in the literature suggesting that it is possible to construct a prediction rule from only a few genes such that it has a negligible prediction error rate. However, in these results the test error or the leave-one-out cross-validated error is calculated without allowance for the selection bias. There is no allowance because the rule is either tested on tissue samples that were used in the first instance to select the genes being used in the rule or because the cross-validation of the rule is not external to the selection process; that is, gene selection is not performed in training the rule at each stage of the cross-validation process. We describe how in practice the selection bias can be assessed and corrected for by either performing a cross-validation or applying the bootstrap external to the selection process. We recommend using 10-fold rather than leave-one-out cross-validation, and concerning the bootstrap, we suggest using the so-called .632+ bootstrap error estimate designed to handle overfitted prediction rules. Using two published data sets, we demonstrate that when correction is made for the selection bias, the cross-validated error is no longer zero for a subset of only a few genes.

We consider the semi-supervised learning problem, where a decision rule is to be learned from labeled and unlabeled data. In this framework, we motivate minimum entropy regularization, which enables to incorporate unlabeled data in the standard supervised learning. Our approach in-cludes other approaches to the semi-supervised problem as particular or limiting cases. A series of experiments illustrates that the proposed solu-tion benefits from unlabeled data. The method challenges mixture mod-els when the data are sampled from the distribution class spanned by the generative model. The performances are definitely in favor of minimum entropy regularization when generative models are misspecified, and the weighting of unlabeled data provides robustness to the violation of the “cluster assumption”. Finally, we also illustrate that the method can also be far superior to manifold learning in high dimension spaces. 1

This article presents a vision for future unmanned aerial vehicles (UAV)-assisted disaster management, considering the holistic functions of disaster prediction, assessment, and response. Here, UAVs not only survey the affected area but also assist in establishing vital wireless communication links between the survivors and nearest available cellular infrastructure. A perspective of different classes of geophysical, climate-induced, and meteorological disasters based on the extent of interaction between the UAV and terrestrially deployed wireless sensors is presented in this work, with suitable network architectures designed for each of these cases. The authors outline unique research challenges and possible solutions for maintaining connected aerial meshes for handoff between UAVs and for systems-specific, security- and energy-related issues. This article is part of a special issue on drones.

In this paper, we present a controller design and its implementation on a mini rotorcraft having four rotors. The dynamic model of the four-rotor rotorcraft is obtained via a Lagrange approach. The proposed controller is based on Lyapunov analysis using a nested saturation algorithm. The global stability analysis of the closed-loop system is presented. Real-time experiments show that the controller is able to perform autonomously the tasks of taking off, hovering, and landing.

This paper gives a broad overview of the stability and control of time-delay systems. Emphasis is on the more recent progress and engineering applications. Examples of practical problems, mathematical descriptions, stability and performance analysis, and feedback control are discussed.

Nowadays, there is a trend to design complex, yet secure systems. In this context, the Trusted Execution Environment (TEE) was designed to enrich the previously defined trusted platforms. TEE is commonly known as an isolated processing environment in which applications can be securely executed irrespective of the rest of the system. However, TEE still lacks a precise definition as well as representative building blocks that systematize its design. Existing definitions of TEE are largely inconsistent and unspecific, which leads to confusion in the use of the term and its differentiation from related concepts, such as secure execution environment (SEE). In this paper, we propose a precise definition of TEE and analyze its core properties. Furthermore, we discuss important concepts related to TEE, such as trust and formal verification. We give a short survey on the existing academic and industrial ARM TrustZone-based TEE, and compare them using our proposed definition. Finally, we discuss some known attacks on deployed TEE as well as its wide use to guarantee security in diverse applications.

This paper presents the control of an underactuated two-link robot called the Pendubot. We propose a controller for swinging the linkage and raise it to its uppermost unstable equilibrium position. The balancing control is based on an energy approach and the passivity properties of the system.

International audience

International audience

Predicting other traffic participants trajectories is a crucial task for an autonomous vehicle, in order to avoid collisions on its planned trajectory. It is also necessary for many Advanced Driver Assistance Systems, where the ego-vehicle's trajectory has to be predicted too. Even if trajectory prediction is not a deterministic task, it is possible to point out the most likely trajectory. This paper presents a new trajectory prediction method which combines a trajectory prediction based on Constant Yaw Rate and Acceleration motion model and a trajectory prediction based on maneuver recognition. It takes benefit on the accuracy of both predictions respectively a short-term and long-term. The defined Maneuver Recognition Module selects the current maneuver from a predefined set by comparing the center lines of the road's lanes to a local curvilinear model of the path of the vehicle. The overall approach was tested on prerecorded human real driving data and results show that the Maneuver Recognition Module has a high success rate and that the final trajectory prediction has a better accuracy.

The fast-paced development of Unmanned Aerial Vehicles (UAVs) and their use in different domains, opens a new paradigm on their use in natural disaster management. In UAV-assisted disaster management applications, UAVs not only survey the affected area but also assist in establishing the communication network between the disaster survivors, rescue teams and nearest available cellular infrastructure. This paper identifies main disaster management applications of UAV networks and discusses open research issues related to UAV-assisted disaster management.

Many data such as social networks, movie preferences or knowledge bases are multi-relational, in that they describe multiple relations between entities. While there is a large body of work focused on modeling these data, modeling these multiple types of relations jointly remains challenging. Further, existing approaches tend to breakdown when the number of these types grows. In this paper, we propose a method for modeling large multi-relational datasets, with possibly thousands of relations. Our model is based on a bilinear structure, which captures various orders of interaction of the data, and also shares sparse latent factors across different relations. We illustrate the performance of our approach on standard tensor-factorization datasets where we attain, or outperform, state-of-the-art results. Finally, a NLP application demonstrates our scalability and the ability of our model to learn efficient and semantically meaningful verb representations. 1

Open-text (or open-domain) semantic parsers are designed to interpret any statement in natural language by inferring a corresponding meaning representation (MR). Unfortunately, large scale systems cannot be easily machine-learned due to lack of directly supervised data. We propose here a method that learns to assign MRs to a wide range of text (using a dictionary of more than 70,000 words, which are mapped to more than 40,000 entities) thanks to a training scheme that combines learning from knowledge bases (WordNet and ConceptNet) with learning from raw text. The model jointly learns representations of words, entities and MRs via a multi-task training process operating on these diverse sources of data. Hence, the system ends up providing methods for knowledge acquisition and word-sense disambiguation within the context of semantic parsing in a single elegant framework. Experiments on these various tasks indicate the promise of the approach. 1

Adaptive control provides techniques for adjusting control parameters in real time to maintain system performance despite unknown or changing process parameters. These methods use real data to tune controllers and adjust plant models or controller parameters. The field has progressed significantly since the 1970s, helped by digital computers. Early applications offered essential feedback, and theoretical advances solved many basic problems. This book comprehensively treats adaptive control, guiding readers from basic problems to analytical solutions with practical applications. Presenting a unified view is challenging due to various design steps and applications. However, a coherent presentation of basic techniques is now possible. The book uses a discrete-time approach to reflect the role of digital computers and shares practical experiences and understanding of different control designs. Mathematical aspects of synthesizing and analyzing algorithms are emphasized, though they alone may not solve practical problems. The book includes applications of control techniques but stresses that a solid mathematical understanding is crucial for creatively applying them to new challenges. Mathematical synthesis and analysis are highlighted, but they must be supplemented with practical problem-solving and algorithm modifications for specific applications.

The problem of the global uniform asymptotic stabilization by bounded feedback of a chain of integrators with a delay in the input is solved. No limitation on the size of the delay is imposed. To validate the approach, a third-order example is presented.

Ouvrage publié avec l'aide du Ministère des affaires étrangères, direction de la coopération scientifique et technique. AVERTISSEMENT Le livre publié en 1995 chez MASSON (EAN13 : 9782225849237) est épuisé. Cette version pdf est une version élaborée à partie de la version préliminaire transmise à l'éditeur. La mise en page est légèrement différente de celle du livre. Malheureusement quelques figures de l'annexe C ont été perdues.

This note addresses the output feedback stabilization problem of a chain of integrators using multiple delays. We shall prove that either n distinct delays or a proportional+delay compensator with n-1 distinct delays are sufficient to stabilize a chain including n integrators. We present two different approaches. Both are constructive and rely on frequency-domain techniques: on a derivative feedback approximation idea, and a pole placement idea, respectively. An illustrative example (triple integrator) is presented.