Guiming Luo

Large Kernel Matters — Improve Semantic Segmentation by Global Convolutional Network

One of recent trends [31, 32, 14] in network architecture design is stacking small filters (e.g., 1x1 or 3x3) in the entire network because the stacked small filters is more efficient than a large kernel, given the same computational complexity. However, in the field of semantic segmentation, where we need to perform dense per-pixel prediction, we find that the large kernel (and effective receptive field) plays an important role when we have to perform the classification and localization tasks simultaneously. Following our design principle, we propose a Global Convolutional Network to address both the classification and localization issues for the semantic segmentation. We also suggest a residual-based boundary refinement to further refine the object boundaries. Our approach achieves state-of-art performance on two public benchmarks and significantly outperforms previous results, 82.2% (vs 80.2%) on PASCAL VOC 2012 dataset and 76.9% (vs 71.8%) on Cityscapes dataset.

Argument Ranking with Categoriser Function

Recently, ranking-based semantics is proposed to rank-order arguments from the most acceptable to the weakest one(s), which provides a graded assessment to arguments. In general, the ranking on arguments is derived from the strength values of the arguments. Categoriser function is a common approach that assigns a strength value to a tree of arguments. When it encounters an argument system with cycles, then the categoriser strength is the solution of the non-linear equations. However, there is no detail about the existence and uniqueness of the solution, and how to find the solution (if exists). In this paper, we will cope with these issues via fixed point technique. In addition, we define the categoriser-based ranking semantics in light of categoriser strength, and investigate some general properties of it. Finally, the semantics is shown to satisfy some of the axioms that a ranking-based semantics should satisfy.

An Algorithm for Calculating the Satisfiability Degree

In the real world, there are situations of uncertainty that the classical logics are not capable to represent, consequently some non-classical logics emerged to compensate for the deficiency of the classical logics of expressing uncertainty. Satisfiability degree was presented as a new means of representing uncertainty. In this talk some properties of the satisfiability degree are viewed. An algorithm is proposed for computing the satisfiability degree of any proposition, which is developed by using XOBDD, an extension of OBDD.

Backtracking Search Algorithm for Satisfiability Degree Calculation

Satisfiability degree can compensate for the deficiency of classical logics in representing uncertainty. As the calculation of the satisfiability degree is an NP-complete problem, it is necessary to construct some efficient algorithm. In this paper, an algorithm for computing the satisfiability degree of arbitrary propositional formula is proposed. It refers to and modifies the backtracking search algorithm used in the Boolean satisfiability problem (SAT), making optimizations by using heuristic strategy and intelligent analysis. Experimental results compare the time consumed by the basic enumeration algorithm with this algorithm, indicating that this algorithm is extremely effective for large formulas.

Short term power load prediction with knowledge transfer

A novel transfer learning method is proposed in this paper to solve the power load forecast problems in the smart grid. Prediction errors of the target tasks can be greatly reduced by utilizing the knowledge transferred from the source tasks. In this work, a source task selection algorithm is developed and the transfer learning model based on Gaussian process is constructed. Negative knowledge transfers are avoided compared with the previous works, and therefore the prediction accuracies are greatly improved. In addition, a fast inference algorithm is developed to accelerate the prediction steps. The results of the experiments with real world data are illustrated.

Frisch scheme identification for Errors-in-Variables systems

This paper considers the problem of identification of dynamic Errors-in-Variables (EIV) systems. Some fatal errors of the well-known Frisch Scheme for EIV identification have been presented, and on the basis of it an improved recursive algorithm is proposed. The new algorithm can estimate both the system parameters and the noise variance with higher accuracy and computational efficiency. Simulations illustrate the theoretical results.

Some Supplementaries to the Counting Semantics for Abstract Argumentation

Dungs abstract argumentation framework consists of a set of interacting arguments and a series of semantics for evaluating them. Those semantics partition the powerset of the set of arguments into two classes: extensions and non-extensions. In order to reason with a specific semantics, one needs to take a credulous or skeptical approach, i.e. an argument is eventually accepted, if it is accepted in one or all extensions, respectively. In our previous work [1], we have proposed a novel semantics, called counting semantics, which allows for a more fine-grained assessment to arguments by counting the number of their respective attackers and defenders based on argument graph and argument game. In this paper, we continue our previous work by presenting some supplementaries about how to choose the damaging factor for the counting semantics, and what relationships with some existing approaches, such as Dungs classical semantics, generic gradual valuations. Lastly, an axiomatic perspective on the ranking semantics induced by our counting semantics are presented.

Automatic verification of event-driven control programs: A case study

Execution environment and temporal performance are very important for modeling a real PLC system. They also become two impediments when verifying with model checking. This paper presents a methodology for modeling a PLC system including time and environment. In order to take into account interaction of PLC controller with execution environment, the proposed method extends the modeling method with an event-driven mechanism. The heterogeneous environments are abstracted as concurrent entities and further formalized into state graphs. A timed abstraction method is proposed to deal with real-time features which specified as timer on delay (TON). We have validated our approach on a concrete case study, a controller for steel plate transfer devices, and report on the results obtained for this case study.

A New Recursive Kernel Regression Algorithm and Its Application in Ultra-short Time Power Load Forecasting

Abstract A new kernel regression algorithm is introduced in this work. The algorithm combines the kernel method and the recursive least squares method, and it is especially useful for on-line time series forecasting and non-linear systems. Under the interruption of colored noises, this algorithm performed better than existing algorithms of the same kind. We validated the algorithms accuracy and time cost using both numerical simulations and real-world experimental data.

Proposition matrix search algorithm for satisfiability degree computation

Satisfiability degree is used as a new means of representing uncertainty. It is able to express the extent of a system satisfying some property. How to compute the satisfiability degree is a critically problem. Although this paper has proved the computation for the satisfiability degree has the exponential worst case complexity, the proposition matrix is proposed to construct the proposition matrix search algorithm. It always chooses the high frequency proposition to simplify the old formula to obtain two new formulae, and the satisfiability degree of the old formula is equal to the sum of satisfiability degrees of the two new formulae. If the new formulae can be judged as true or false, the algorithm directly computes their satisfiability degree; else their satisfiability degree is recursively computed as the old formula. The proposition matrix search algorithm uses a truth detector to detect the truth value of a proposition so that computation times can be reduced significantly. Experimental results show it is more effective than the basic enumeration algorithm and the backtracking search algorithm.