Exploratory learning structures in artificial cognitive systems

The major goal of the COSPAL project is to develop an artificial cognitive system architecture, with the ability to autonomously extend its capabilities. Exploratory learning is one strategy that allows an extension of competences as provided by the environment of the system. Whereas classical learning methods aim at best for a parametric generalization, i.e., concluding from a number of samples of a problem class to the problem class itself, exploration aims at applying acquired competences to a new problem class, and to apply generalization on a conceptual level, resulting in new models. Incremental or online learning is a crucial requirement to perform exploratory learning.In the COSPAL project, we mainly investigate reinforcement-type learning methods for exploratory learning, and in this paper we focus on the organization of cognitive systems for efficient operation. Learning is used over the entire system. It is organized in the form of four nested loops, where the outermost loop reflects the user-reinforcement-feedback loop, the intermediate two loops switch between different solution modes at symbolic respectively sub-symbolic level, and the innermost loop performs the acquired competences in terms of perception–action cycles. We present a system diagram which explains this process in more detail.We discuss the learning strategy in terms of learning scenarios provided by the user. This interaction between user (’teacher’) and system is a major difference to classical robotics systems, where the system designer places his world model into the system. We believe that this is the key to extendable robust system behavior and successful interaction of humans and artificial cognitive systems.We furthermore address the issue of bootstrapping the system, and, in particular, the visual recognition module. We give some more in-depth details about our recognition method and how feedback from higher levels is implemented. The described system is however work in progress and no final results are available yet. The available preliminary results that we have achieved so far, clearly point towards a successful proof of the architecture concept.     

Worst-case response time analysis of real-time tasks under fixed-priority scheduling with deferred preemption

Fixed-priority scheduling with deferred preemption (FPDS) has been proposed in the literature as a viable alternative to fixed-priority pre-emptive scheduling (FPPS), that obviates the need for non-trivial resource access protocols and reduces the cost of arbitrary preemptions. This paper shows that existing worst-case response time analysis of hard real-time tasks under FPDS, arbitrary phasing and relative deadlines at most equal to periods is pessimistic and/or optimistic. The same problem also arises for fixed-priority non-pre-emptive scheduling (FPNS), being a special case of FPDS. This paper provides a revised analysis, resolving the problems with the existing approaches. The analysis is based on known concepts of critical instant and busy period for FPPS. To accommodate for our scheduling model for FPDS, we need to slightly modify existing definitions of these concepts. The analysis assumes a continuous scheduling model, which is based on a partitioning of the timeline in a set of non-empty, right semi-open intervals. It is shown that the critical instant, longest busy period, and worst-case response time for a task are suprema rather than maxima for all tasks, except for the lowest priority task. Hence, that instant, period, and response time cannot be assumed for any task, except for the lowest priority task. Moreover, it is shown that the analysis is not uniform for all tasks, i.e. the analysis for the lowest priority task differs from the analysis of the other tasks. These anomalies for the lowest priority task are an immediate consequence of the fact that only the lowest priority task cannot be blocked. To build on earlier work, the worst-case response time analysis for FPDS is expressed in terms of known worst-case analysis results for FPPS. The paper includes pessimistic variants of the analysis, which are uniform for all tasks, illustrates the revised analysis for an advanced model for FPDS, where tasks are structured as flow graphs of subjobs rather than sequences, and shows that our analysis is sustainable.     

General Adaptive Neighborhood Choquet Image Filtering

A novel framework entitled General Adaptive Neighborhood Image Processing (GANIP) has been recently introduced in order to propose an original image representation and mathematical structure for adaptive image processing and analysis. The central idea is based on the key notion of adaptivity which is simultaneously associated with the analyzing scales, the spatial structures and the intensity values of the image to be addressed. In this paper, the GANIP framework is briefly exposed and particularly studied in the context of Choquet filtering (using fuzzy measures), which generalizes a large class of image filters. The resulting spatially-adaptive operators are studied with respect to the general GANIP framework and illustrated in both the biomedical and materials application areas. In addition, the proposed GAN-based filters are practically applied and compared to several other denoising methods through experiments on image restoration, showing a high performance of the GAN-based Choquet filters.

Discourses on age: confronting disputed concepts by means of dialogue

The article explores approaches to discourses concerning age, with different agendas and national contexts. The Dialogue Seminar Method is introduced, as a means of facilitating reflection and access to tacit knowledge. Democratic dialogue requires orchestration, and enables horizontal communication and collective reflection.

---

Globally conservative high-order filters for large-eddy simulation and computational aero-acoustics

In computational aero-acoustics, large-eddy simulations (LES) or direct numerical simulations (DNS) are often employed for flow computations in the source region. As part of the numerical implementation or required modeling, explicit spatial filters are frequently employed. For instance, in LES spatial filters are employed in the formulation of various subgrid-scale (SGS) models such as the dynamic model or the variational multi-scale (VMS) Smagorinsky model; both in LES or DNS, spatial high-pass filters are often used to remove undesired grid-to-grid oscillations. Though these type of spatial filters adhere to local accuracy requirements, in practice, they often destroy global conservation properties in the presence of non-periodic boundaries conditions. This leads to the incorrect prediction of the flow properties near hard boundaries, such as walls. In the current work, we present globally conservative high-order accurate filters, which combine traditional filters at the internal points with one-sided conservative filters near the wall boundary. We test these filters to remove grid-to-grid oscillations both in a channel-flow case and in 2D cavity flow. We find that the use of a non-conservative filter leads to erroneous predictions of the skin friction in channel flows up to 30%. In the cavity-flow simulations, the use of non-conservative filters to remove grid-to-grid oscillations leads to important shifts in the Strouhal number of the dominant mode, and a change of the flow pattern inside the cavity. In all cases, the use of conservative high-order filter formulations to remove grid-to-grid oscillations lead to very satisfactory results. Finally, in our channel-flow test case, we also illustrate the importance of using conservative filters for the formulation of the VMS Smagorinsky model.     

Predicting the unexpected

Dagstuhl seminar no. 10102 on discrete event logistic systems recognized a network of persistent models to be a “Grand Challenge.” Such on-line model network will offer an infrastructure that facilitates the management of logistic operations. This ambition to create a network of persistent models implies a radical shift for model design activities as the objective is an infrastructure rather than application-specific solutions. In particular, model developers can no longer assume that they know what their model will be used for. It is no longer possible to design for the expected.This paper presents insights in model development and design in the absence of precise knowledge concerning a model's usage. Basically, model developers may solely rely on the presence of the real-world counterpart mirrored by their model and a general idea about the nature of the application (e.g. coordination of logistic operations). When the invariants of their real-world counterpart suffice for models to be valid, these models become reusable and integrate-able. As these models remain valid under a wide range of situations, they become multi-purpose and durable resources rather than single-purpose short-lived components or legacy, which is even worse.Moreover and more specifically, the paper describes how to build models that allow their users to generate predictions in unexpected situations and atypical conditions. Referring to previous work, the paper concisely discusses how these predictions can be generated starting from the models. This prediction-generating technology is currently being transferred into an industrial MES.     

An empirical evaluation of the comprehensibility of decision table, tree and rule based predictive models

An important objective of data mining is the development of predictive models. Based on a number of observations, a model is constructed that allows the analysts to provide classifications or predictions for new observations. Currently, most research focuses on improving the accuracy or precision of these models and comparatively little research has been undertaken to increase their comprehensibility to the analyst or end-user. This is mainly due to the subjective nature of ‘comprehensibility’, which depends on many factors outside the model, such as the user's experience and his/her prior knowledge. Despite this influence of the observer, some representation formats are generally considered to be more easily interpretable than others. In this paper, an empirical study is presented which investigates the suitability of a number of alternative representation formats for classification when interpretability is a key requirement. The formats under consideration are decision tables, (binary) decision trees, propositional rules, and oblique rules. An end-user experiment was designed to test the accuracy, response time, and answer confidence for a set of problem-solving tasks involving the former representations. Analysis of the results reveals that decision tables perform significantly better on all three criteria, while post-test voting also reveals a clear preference of users for decision tables in terms of ease of use.     

3D model alignment based on minimum projection area

3D model alignment is an important step for applications such as 3D model retrieval and 3D model recognition. In this paper, we propose a novel Minimum Projection Area-based (MPA) alignment method for pose normalization. Our method finds three principal axes to align a model: the first principal axis gives the minimum projection area when we perform an orthographic projection of the model in the direction parallel to this axis, the second axis is perpendicular to the first axis and gives the minimum projection area, and the third axis is the cross product of the first two axes. We devise an optimization method based on Particle Swarm Optimization to efficiently find the axis with minimum projection area. For application in retrieval, we further perform axis ordering and orientation in order to align similar models in similar poses. We have tested MPA on several standard databases which include rigid/non-rigid and open/watertight models. Experimental results demonstrate that MPA has a good performance in finding alignment axes which are parallel to the ideal canonical coordinate frame of models and aligning similar models in similar poses under different conditions such as model variations, noise, and initial poses. In addition, it achieves a better 3D model retrieval performance than several commonly used approaches such as CPCA, NPCA, and PCA.     

On partitioning a graph into two connected subgraphs

Suppose a graph G is given with two vertex-disjoint sets of vertices Z₁ and Z₂. Can we partition the remaining vertices of G such that we obtain two connected vertex-disjoint subgraphs of G that contain Z₁ and Z₂, respectively? This problem is known as the 2-Disjoint Connected Subgraphs problem. It is already NP-complete for the class of n-vertex graphs G=(V,E) in which Z₁ and Z₂ each contain a connected set that dominates all vertices in V?(Z₁∪Z₂). We present an O^∗(1.2051ⁿ) time algorithm that solves it for this graph class. As a consequence, we can also solve this problem in O^∗(1.2051ⁿ) time for the classes of n-vertex P₆-free graphs and split graphs. This is an improvement upon a recent O^∗(1.5790ⁿ) time algorithm for these two classes. Our approach translates the problem to a generalized version of hypergraph 2-coloring and combines inclusion/exclusion with measure and conquer.