Fault detection and accommodation via. neural network and variable structure control

This paper proposes a novel idea that classifies faults into two different kinds： serious faults and small faults, and treats them with different strategies respectively. A kind of artificial neural network （ANN） is proposed for detecting serious faults, and variable structure （VS） model-following control is constructed for accommodating small faults. The proposed framework takes both advantages of qualitative way and quantitative way of fault detection and accommodation. Moreover, the uncertainty case is investigated and the VS controller is modified. Simulation results of a remotely piloted aircraft with control actuator failures illustrate the performance of the developed algorithm.     

Single machine scheduling models with deterioration and learning: handling precedence constraints via priority generation

We consider various single machine scheduling problems in which the processing time of a job depends either on its position in a processing sequence or on its start time. We focus on problems of minimizing the makespan or the sum of (weighted) completion times of the jobs. In many situations we show that the objective function is priority-generating, and therefore the corresponding scheduling problem under series-parallel precedence constraints is polynomially solvable. In other situations we provide counter-examples that show that the objective function is not priority-generating.     

Boosting learning and inference in Markov logic through metaheuristics

Markov Logic (ML) combines Markov networks (MNs) and first-order logic by attaching weights to first-order formulas and using these as templates for features of MNs. State-of-the-art structure learning algorithms in ML maximize the likelihood of a database by performing a greedy search in the space of structures. This can lead to suboptimal results because of the incapability of these approaches to escape local optima. Moreover, due to the combinatorially explosive space of potential candidates these methods are computationally prohibitive. We propose a novel algorithm for structure learning in ML, based on the Iterated Local Search (ILS) metaheuristic that explores the space of structures through a biased sampling of the set of local optima. We show through real-world experiments that the algorithm improves accuracy and learning time over the state-of-the-art algorithms. On the other side MAP and conditional inference for ML are hard computational tasks. This paper presents two algorithms for these tasks based on the Iterated Robust Tabu Search (IRoTS) metaheuristic. The first algorithm performs MAP inference and we show through extensive experiments that it improves over the state-of-the-art algorithm in terms of solution quality and inference time. The second algorithm combines IRoTS steps with simulated annealing steps for conditional inference and we show through experiments that it is faster than the current state-of-the-art algorithm maintaining the same inference quality.     

Fault detection in autonomous robots based on fault injection and learning

In this paper, we study a new approach to fault detection for autonomous robots. Our hypothesis is that hardware faults change the flow of sensory data and the actions performed by the control program. By detecting these changes, the presence of faults can be inferred. In order to test our hypothesis, we collect data from three different tasks performed by real robots. During a number of training runs, we record sensory data from the robots while they are operating normally and after a fault has been injected. We use back-propagation neural networks to synthesize fault detection components based on the data collected in the training runs. We evaluate the performance of the trained fault detectors in terms of number of false positives and time it takes to detect a fault. The results show that good fault detectors can be obtained. We extend the set of possible faults and go on to show that a single fault detector can be trained to detect several faults in both a robot’s sensors and actuators. We show that fault detectors can be synthesized that are robust to variations in the task, and we show how a fault detector can be trained to allow one robot to detect faults that occur in another robot.

New Possibilities in Image Diffusion and Sharpening via High-Order Sobolev Gradient Flows

We study the use of high-order Sobolev gradients for PDE-based image smoothing and sharpening, extending our previous work on this problem. In particular, we study the gradient descent equation on the heat equation energy functional obtained by modifying the usual metric on the space of images, which is the L ² metric, to a weighted H ^k Sobolev metric. We present existence and uniqueness results which show that the Sobolev diffusion PDE are well-posed both in the forward and backward direction. Furthermore, we perform a Fourier analysis on the scale space generated by the Sobolev PDE and show that as the order of the Sobolev metric tends to infinity, the Sobolev gradients converge to a Gaussian smoothed L ² gradient. We then present experimental results which exploit the theoretical stability results by applying the various Sobolev gradient flows in the backward direction for image sharpening effects. Furthermore, we show that as the Sobolev order is increased, the sharpening effects become more global in nature and more immune to noise.     

Particle swarm optimizer for variable weighting in clustering high-dimensional data

In this paper, we present a particle swarm optimizer (PSO) to solve the variable weighting problem in projected clustering of high-dimensional data. Many subspace clustering algorithms fail to yield good cluster quality because they do not employ an efficient search strategy. In this paper, we are interested in soft projected clustering. We design a suitable k-means objective weighting function, in which a change of variable weights is exponentially reflected. We also transform the original constrained variable weighting problem into a problem with bound constraints, using a normalized representation of variable weights, and we utilize a particle swarm optimizer to minimize the objective function in order to search for global optima to the variable weighting problem in clustering. Our experimental results on both synthetic and real data show that the proposed algorithm greatly improves cluster quality. In addition, the results of the new algorithm are much less dependent on the initial cluster centroids. In an application to text clustering, we show that the algorithm can be easily adapted to other similarity measures, such as the extended Jaccard coefficient for text data, and can be very effective.     

Estimating the feasibility of transition paths in extended finite state machines

There has been significant interest in automating testing on the basis of an extended finite state machine (EFSM) model of the required behaviour of the implementation under test (IUT). Many test criteria require that certain parts of the EFSM are executed. For example, we may want to execute every transition of the EFSM. In order to find a test suite (set of input sequences) that achieves this we might first derive a set of paths through the EFSM that satisfy the criterion using, for example, algorithms from graph theory. We then attempt to produce input sequences that trigger these paths. Unfortunately, however, the EFSM might have infeasible paths and the problem of determining whether a path is feasible is generally undecidable. This paper describes an approach in which a fitness function is used to estimate how easy it is to find an input sequence to trigger a given path through an EFSM. Such a fitness function could be used in a search-based approach in which we search for a path with good fitness that achieves a test objective, such as executing a particular transition, and then search for an input sequence that triggers the path. If this second search fails then we search for another path with good fitness and repeat the process. We give a computationally inexpensive approach (fitness function) that estimates the feasibility of a path. In order to evaluate this fitness function we compared the fitness of a path with the ease with which an input sequence can be produced using search to trigger the path and we used random sampling in order to estimate this. The empirical evidence suggests that a reasonably good correlation (0.72 and 0.62) exists between the fitness of a path, produced using the proposed fitness function, and an estimate of the ease with which we can randomly generate an input sequence to trigger the path.     

Adaptive-resolution reinforcement learning with polynomial exploration in deterministic domains

We propose a model-based learning algorithm, the Adaptive-resolution Reinforcement Learning (ARL) algorithm, that aims to solve the online, continuous state space reinforcement learning problem in a deterministic domain. Our goal is to combine adaptive-resolution approximation schemes with efficient exploration in order to obtain polynomial learning rates. The proposed algorithm uses an adaptive approximation of the optimal value function using kernel-based averaging, going from coarse to fine kernel-based representation of the state space, which enables us to use finer resolution in the “important” areas of the state space, and coarser resolution elsewhere. We consider an online learning approach, in which we discover these important areas online, using an uncertainty intervals exploration technique. In addition, we introduce an incremental variant of the ARL (IARL), which is a more practical version of the original algorithm with reduced computational complexity at each stage. Polynomial learning rates in terms of mistake bound (in a PAC framework) are established for these algorithms, under appropriate continuity assumptions.     

Learning noisy linear classifiers via adaptive and selective sampling

We introduce efficient margin-based algorithms for selective sampling and filtering in binary classification tasks. Experiments on real-world textual data reveal that our algorithms perform significantly better than popular and similarly efficient competitors. Using the so-called Mammen-Tsybakov low noise condition to parametrize the instance distribution, and assuming linear label noise, we show bounds on the convergence rate to the Bayes risk of a weaker adaptive variant of our selective sampler. Our analysis reveals that, excluding logarithmic factors, the average risk of this adaptive sampler converges to the Bayes risk at rate N ^{−(1+α)(2+α)/2(3+α)} where N denotes the number of queried labels, and α>0 is the exponent in the low noise condition. For all $\alpha>\sqrt{3}-1\approx0.73$\alpha>\sqrt{3}-1\approx0.73 this convergence rate is asymptotically faster than the rate N ^{−(1+α)/(2+α)} achieved by the fully supervised version of the base selective sampler, which queries all labels. Moreover, for α→∞ (hard margin condition) the gap between the semi- and fully-supervised rates becomes exponential.     

Sense and sidedness in the graphics pipeline via a passage through a separable space

Computer graphics is ostensibly based on projective geometry. The graphics pipeline—the sequence of functions applied to 3D geometric primitives to determine a 2D image—is described in the graphics literature as taking the primitives from Euclidean to projective space, and then back to Euclidean space. This is a weak foundation for computer graphics. An instructor is at a loss: one day entering the classroom and invoking the established and venerable theory of projective geometry while asserting that projective spaces are not separable, and then entering the classroom the following week to tell the students that the standard graphics pipeline performs clipping not in Euclidean, but in projective space—precisely the operation (deciding sidedness, which depends on separability) that was deemed nonsensical. But there is no need to present Blinn and Newell’s algorithm (Comput. Graph. 12, 245–251, 1978; Commun. ACM 17, 32–42, 1974)—the crucial clipping step in the graphics pipeline and, perhaps, the most original knowledge a student learns in a fourth-year computer graphics class—as a clever trick that just works. Jorge Stolfi described in 1991 oriented projective geometry. By declaring the two vectors and distinct, Blinn and Newell were already unknowingly working in oriented projective space. This paper presents the graphics pipeline on this stronger foundation.