On Structured Output Training: Hard Cases and an Efficient Alternative p. 7 Spares Kernel SVMs via Cutting-Plane Training p. 8 Hybrid Least-Squares Algorithms for Approximate Policy Evaluation p. 9 A Self-training Approach to Cost Sensitive Uncertainty Sampling p. 10 Learning Multi-linear Representations of Distributions for Efficient Inference p. 11 Cost-Sensitive Learning Based on Bregman Div...

Parallel graph partitioning is a di cult issue, because the best sequential graph partitioning methods known to date are based on iterative local optimization algorithms that do not parallelize nor scale well. On the other hand, evolutionary algorithms are highly parallel and scalable, but converge very slowly as problem size increases. This paper presents methods that can be used to reduce pro...

In the paper, we present a new approach to multi-way Blind Source Separation (BSS) and corresponding 3D tensor factorization that has many potential applications in neuroscience and multi-sensory or multidimensional data analysis, and neural sparse coding. We propose to use a set of local cost functions with flexible penalty and regularization terms whose simultaneous or sequential (one by one)...

In this paper, we study the scheduling problem of real-time disk requests in multi-disk systems, such as RAID-0. We first propose a multi-disk scheduling algorithm, called Least-Remaining-Request-Size-First (LRSF), to improve soft real-time performance of I/O systems. LRSF may be integrated with different real-time/non-real-time single-disk scheduling algorithms, such as SATF and SSEDV, adopted...

In the case of convection dominated problems, multi-grid methods require an appropriate smoothing to ensure robustness. As a first approach we discuss a Gauß-Seidel smoothing with a correct numbering of the unknowns and if necessary a special block partitioning. Numerical experiments show that, in the case of general convection directions, the multi-grid algorithms obtained in this way have the...

Rapid visible/near-infrared (VNIR) hyperspectral imaging methods, employing both a single waveband algorithm and multi-spectral algorithms, were developed in order to discrimination between sound and discolored lettuce. Reflectance spectra for sound and discolored lettuce surfaces were extracted from hyperspectral reflectance images obtained in the 400-1000 nm wavelength range. The optimal wave...

Data locality and synchronization overhead are two important factors that affect the performance of applications on multiprocessors. Loop fusion is an effective way for reducing synchronization and improving data locality. Traditional fusion techniques, however, either can not address the case when fusion-preventing dependencies exist in nested loops, or can not achieve good parallelism after f...

The current study proposes a two-step approach for pore facies characterization in the carbonate reservoirs with an example from the Kangan and Dalanformations in the South Pars gas field. In the first step, pore facies were determined based on Mercury Injection Capillary Pressure (MICP) data incorporation with the Hierarchical Clustering Analysis (HCA) method. In the next step, polynomial meta...

– Natural multi-agent systems like ant colonies prove that decentralized systems based on simple constituent parts can exhibit advanced global behavior that has important advantages over explicit central control in others systems. Unfortunately, it is difficult to design a collection of individual components and their local interaction in a way that will give rise to a successful global behavio...