An Object-Oriented Modeling and Simulation Environment for Reactive Systems Development

This paper discusses an architecture designed to provide support for the development of state transition models for an object-oriented distributed environment. The state transition models can be constructed and specified hierarchically as well as derived into new classes of objects through the use of inheritance. A new concept called exit-safe states is introduced to assist in the specification of hierarchical state transition models. A graphical monitor to analyze the system at run time has been developed.

Intelligent vehicles need to detect new classes of traffic objects while keeping the performance of old ones. Deep convolution neural network (DCNN) based detector has shown superior performance, however, DCNN is ill-equipped for incremental learning, i.e., a DCNN based vehicle detector trained on traffic sign dataset will catastrophic forget how to detect vehicles. In this paper, we propose a novel method to alleviate this problem, our key insight is that the original class of objects also appears in new task data, by utilizing these objects, our method effectively keeps the detection accuracy of original models while incremental learning to detect new classes of objects. Detailed experiments on PASCAL VOC dataset and TSD-max database verified the effectiveness of our method.

The debate over the demotion of Pluto to a “dwarf planet” in Prague at the IAU General Assembly in 2006 showed that classification is far from a boring subject, and in fact opens a window on a little-known aspect of the history of astronomy over the last 400 years. How does an astronomer know when he or she has discovered a new class of astronomical object? Who decides if a dwarf planet, or a quasar or a pulsar, is a new class of object? To put it more broadly, how does an astronomer know if he or she has discovered something new, especially to the extent that it is declared a totally new class of object? This chapter begins with the Pluto debate, but quickly broadens out to the discovery of other new classes of objects over the last 400 years of telescopic astronomy. Classification is an open-ended process. Even as the vote was taken in Prague, NASA’s New Horizons spacecraft was speeding toward the edge of the Solar System, some three billion miles away, to explore the nature of the object so much in dispute. The discoveries made there revitalized the debate over its classification status, as did the arrival of NASA’s Dawn spacecraft at another newly designated dwarf planet, the asteroid Ceres.

The ability to react on and adapt to challenges of the working environment is important and desired characteristic of modern knowledge-based systems. Knowledge extraction from different data sources and representation of heterogeneous knowledge are illustrative examples of such challenges. For effective solving of these tasks knowledge-based systems should be able to generate (create) new classes of objects dynamically, when required. Therefore, the algorithms for dynamic creation of new classes of objects via computing the homogeneous and inhomogeneous intersection of homogeneous and inhomogeneous classes of objects are proposed in the paper. New classes of objects, established via proposed algorithms, define set of types, which determine heterogeneous collections of objects. Proposed approach provides an ability to check the existence of common parts for different classes of objects and to create new classes of objects dynamically, based on previously defined ones. Classes, created using proposed approach, could be integrated within the knowledge base and used for further construction of conceptual hierarchies.

Nowadays, the Convolutional Neural Network is successfully applied to the images object detection. When new classes of object emerges, it is popular to adapt the convolutional neural network based detection model through a retraining process with the new classes of samples. Unfortunately, the adapted model can only detect the new classes of objects, but cannot identify the old classes of objects, which is called catastrophic forgetting, also occurring in incremental classification tasks. Knowledge distillation has achieved good results in incremental learning for classification tasks. Due to the dual tasks within object detection, object classification and location at the same time, a straightforward migration of knowledge distillation method cannot provide a satisfactory result in incremental learning for object detection tasks. Hence, this paper propose a new knowledge distillation for incremental object detection, which introduces a new object detection distillation loss, a loss not only for classification results but also for location results of the predicted bounding boxes, not only for all final detected regions of interest but also for all intermediate regions proposal. Furthermore, to avoid forgetting learned knowledge from old datasets, this paper not only employs hint learning to retain the characteristic information of the initial model, but also innovatively uses confidence loss to extract the confidence information of the initial model. A series of experiment results on the PASCAL VOC 2007 dataset verify the effectiveness of the proposed method.

constructs define new classes of objects. The abstract construct is roughly the inverse of the refine construct. The instantiate, e, -, and infotm serve to include more knowledge about a class of objects, while ~SUZQ.~ is used to represent a class of objects in terms of another class. The instantiate construct records the information that a class of objects can be refined to another class; or, stated another way, a class of objects can be abstracted to another class. The latter could be obtained from the former by reversinq the arguments to instantiate. The use of these constructs is illustrated on structures (see below), a class of objects definable in Tecton, in 151. (Except that abstract was not discussed and refine was called enrich in that paper.) In the next section, we will give examples of some of these constructs; their use is also illustrated in the discussion of the communication network example. We discuss four different types of objects in Tecton which we have found useful in describing different kinds of activities of a complex software system: structures, entities, events, and environ- ments. Some of these types of objects have appeared previously in

With the rapid development of remote sensing technology and the growing demand for applications, the classical deep learning-based object detection model is bottlenecked in processing incremental data, especially in the increasing classes of detected objects. It requires models to sequentially learn new classes of objects based on the current model, while preserving old categories-related knowledge. Existing class-incremental detection methods achieve this goal mainly by constraining the optimization trajectory in the feature of output space. However, these works neglect the case where the previously learned background is a new category to learn, resulting in performance degradation in the new category because of the conflict between remaining the background-related knowledge or updating the background-related knowledge. This paper proposes a novel class-incremental detection method incorporated with the teacher-student architecture and the selective distillation (SDCID) strategy. Specifically, it is the asymmetry architecture, i.e., the teacher network temporarily stores historical knowledge of previously learned objects, and the student network integrates historical knowledge from the teacher network with the newly learned object-related knowledge, respectively. This asymmetry architecture reveals the significance of the distinct representation of history knowledge and new knowledge in incremental detection. Furthermore, SDCID selectively masks the shared subobject of new images to learn and previously learned background, while learning new categories of images and then transfers the classification results of the student model to the background class following the judgment model of the teacher model. This manner avoids interferences between the background category-related knowledge from a teacher model and the learning of other new classes of objects. In addition, we proposed a new incremental learning evaluation metric, C-SP, to comprehensively evaluate the incremental learning stability and plasticity performance. We verified the proposed method on two object detection datasets of remote sensing images, i.e., DIOR and DOTA. The experience results in accuracy and C-SP suggest that the proposed method surpasses existing class-incremental detection methods. We further analyzed the influence of the mask component in our method and the hyper-parameters sensitive to our method.

The integration of new extracted or acquired knowledge into the knowledge base is a crucial task for modern knowledge-based systems, which requires dynamic analysis of the relevance, similarity, and difference of new knowledge. It can be done using special operations defined over the representation structures provided by chosen knowledge model. Within an object-oriented approach, such operations can be implemented in a form of universal exploiters of classes, which adapt and implement for classes the idea of corresponding set-theoretical operations, such as intersection, union, difference, and decomposition into subsets. Therefore corresponding algorithms for implementation of a few forms of universal difference exploiter of classes within such a knowledge representation model as object-oriented dynamic networks are presented in the paper. Developed algorithms can dynamically create new classes of objects via computing the difference of homogeneous and heterogeneous classes as well as the difference of two heterogeneous classes of objects if such a difference exists. The proposed approach provides an opportunity to generate new knowledge structures in a form of classes of objects based on the previously obtained ones. It allows evaluation of the relevance and novelty level of extracted or acquired pieces of knowledge compared with previously obtained ones by computing the difference between them, which can be used for efficient integration of the new knowledge into the knowledge base.

DEOS — A dynamically extendible object-oriented system

This paper presents an object-oriented approach to interactive modelling of geometric objects. The objects are specified by geometric constructions that are built by mimicing the compass-and-ruler manual approach in a direct manipulation graphical interface. The system adopts a programming-by-example paradigm to enrich construction methods incrementally. New constructions can be used to define new classes of objects or new methods for an existing class. Messages exchanged among objects specify geometric relationships among entities. Messages sent at construction time implicitly form a relationship network, which is preserved during subsequent geometric transformations, so that geometric constraints can be satisfied without resorting to numerical methods. The prototype GEObject is implemented under Actor in a Windows 3.0 environment.

Due to the recent dramatic technological advances, infrared interferometry can now be applied to new classes of objects, resulting in exciting new science prospects, for instance, in the area of high-mass star formation. Although extensively studied at various wavelengths, the process through which massive stars form is still only poorly understood. For instance, it has been proposed that massive stars might form like low-mass stars by mass accretion through a circumstellar disk/envelope, or otherwise by coalescence in a dense stellar cluster. After discussing the technological challenges which result from the special properties of these objects, we present first near-infrared interferometric observations, which we obtained on the massive YSO IRAS 13481-6124 using VLTI/AMBER infrared long-baseline interferometry and NTT speckle interferometry. From our extensive data set, we reconstruct a model-independent aperture synthesis image which shows an elongated structure with a size of 13x19 AU, consistent with a disk seen under an inclination of 45 degree. The measured wavelength-dependent visibilities and closure phases allow us to derive the radial disk temperature gradient and to detect a dust-free region inside of 9.5 AU from the star, revealing qualitative and quantitative similarities with the disks observed in low-mass star formation. In complementary mid-infrared Spitzer and sub-millimeter APEX imaging observations we detect two bow shocks and a molecular out ow which are oriented perpendicular to the disk plane and indicate the presence of a bipolar outflow emanating from the inner regions of the system.

Physicists construct ever more energetic particle accelerators in order to probe ever finer details of natural structures. As the energy of the probing particles has gone up, the center of attention has progressed from atoms to nuclei to the structure of such particles as protons and neutrons themselves. With each major increase in energy have come surprises, new classes of objects, new kinds of physical properties, that have forced radical changes in theory. But now, at the level of probing the structure of the so-called elementary particles themselves, some physicists have thought there ought to be a limit. Other physicists are not so sure. The latest step in the parade of energy increases is the 76-billion-electron-volt (GeV) synchrotron at Serpukhov in the Soviet Union. It is more than twice as energetic as anything that operated before it, and when it began experiments about two years ago, the question in physicists' minds was whether experimentation in this range would bring surprises or whether it would merely confirm and extend information already gathered at lower energies. The first Serpukhov experiments have already shown enough surprises to have theorists shaking their heads. Further results are now eagerly awaited. Among the first experiments were so-called total cross-section experiments intended to find out in a general way where things are at this energy range. The cross section reveals the probability that anything at all will happen when an accelerated particle is shot at a target. It depends on a wave that is associated with each particle, the socalled de Broglie wave named for Prince Louis de Broglie, who first suggested its existence in the early 1920's. The de Broglie wave is a way of measuring the probability of a particle's being somewhere in a given volume at a given time; its wavelength determines the area over which a particle's influence is felt, and the probability of some interaction with a target depends on this. At low energies the de Broglie wavelength is larger than the physical size of the particle, but as the energy goes up the wavelength decreases. Eventually a point should come where the wavelength becomes less than the physical size of the particle. At this point the total cross section becomes dependent on the size of the particle and should remain constant if the energy is further increased. But the Serpukhov results are showing that the total cross sections do not approach constant values as fast as current theory indicates they should. A related surprise is that the total cross sections for a particle and that for its antiparticle do not come together at high energies as theory says they should. At low energies the total cross sections for particle and antiparticle differ, since there are natural rules that prohibit antimatter from doing some of the things matter can do. At high energy these rules should lose their effect, and theory says the two cross sections should come to the same value, but experimentally they are not exhibiting the expected behavior. These things seem to be saying that there is something in the nature of the particles that exhibits itself at Serpukhov energies that was not taken into account in present theory, but theorists have so far not come up with an explanation. Other experiments have looked for more surprises. Serpukhov has looked for quarks, the theoretically predicted building blocks out of which the particles are supposed to be made, but has not found them. For several months the accelerator has been shut down for a rearrangement of the experimental hall and beams. When it reopens one of the first experiments will be less direct check of the quark theory than a search for quarks themselves: an investigation of negative mesons. This experiment will be done in collaboration with the CERN laboratory in Geneva as was some of the cross-section work. The quark theory says that all particles are built of either two or three subparticles called quarks, and on this basis makes predictions about their masses and other properties. Previous experiments at CERN have found a series of negatively charged mesons that fit the predictions of the quark theory especially well. The forthcoming

In this paper, similarly to [1,4,17,20,21,29,30], abstract data types are understood as formalized many-sorted theories based on algorithmic languages (e.g. a language of algorithmic logic [2,16] or a language of dynamic logic [11,29]). Operations on data types, leading from (more) primitive types to compound types, are defined in terms of the interpretability theory (cf. Szczerba [25]). The approach proposed here to defining new types accords with the methods of introducing new classes of objects in programming languages like Simula 67, Pascal, Loglan, Ada.

This work is driven by a practical question: corrections of Artificial Intelligence (AI) errors. These corrections should be quick and non-iterative. To solve this problem without modification of a legacy AI system, we propose special ‘external’ devices, correctors. Elementary correctors consist of two parts, a classifier that separates the situations with high risk of error from the situations in which the legacy AI system works well and a new decision that should be recommended for situations with potential errors. Input signals for the correctors can be the inputs of the legacy AI system, its internal signals, and outputs. If the intrinsic dimensionality of data is high enough then the classifiers for correction of small number of errors can be very simple. According to the blessing of dimensionality effects, even simple and robust Fisher’s discriminants can be used for one-shot learning of AI correctors. Stochastic separation theorems provide the mathematical basis for this one-short learning. However, as the number of correctors needed grows, the cluster structure of data becomes important and a new family of stochastic separation theorems is required. We refuse the classical hypothesis of the regularity of the data distribution and assume that the data can have a rich fine-grained structure with many clusters and corresponding peaks in the probability density. New stochastic separation theorems for data with fine-grained structure are formulated and proved. On the basis of these theorems, the multi-correctors for granular data are proposed. The advantages of the multi-corrector technology were demonstrated by examples of correcting errors and learning new classes of objects by a deep convolutional neural network on the CIFAR-10 dataset. The key problems of the non-classical high-dimensional data analysis are reviewed together with the basic preprocessing steps including the correlation transformation, supervised Principal Component Analysis (PCA), semi-supervised PCA, transfer component analysis, and new domain adaptation PCA.

Main sequence stars, like the Sun, are often found to be orbited by\ncircumstellar material that can be categorized into two groups, planets and\ndebris. The latter is made up of asteroids and comets, as well as the dust and\ngas derived from them, which makes debris disks observable in thermal emission\nor scattered light. These disks may persist over Gyrs through steady-state\nevolution and/or may also experience sporadic stirring and major collisional\nbreakups, rendering them atypically bright for brief periods of time. Most\ninterestingly, they provide direct evidence that the physical processes\n(whatever they may be) that act to build large oligarchs from micron-sized dust\ngrains in protoplanetary disks have been successful in a given system, at least\nto the extent of building up a significant planetesimal population comparable\nto that seen in the Solar System's asteroid and Kuiper belts. Such systems are\nprime candidates to host even larger planetary bodies as well. The recent\ngrowth in interest in debris disks has been driven by observational work that\nhas provided statistics, resolved images, detection of gas in debris disks, and\ndiscoveries of new classes of objects. The interpretation of this vast and\nexpanding dataset has necessitated significant advances in debris disk theory,\nnotably in the physics of dust produced in collisional cascades and in the\ninteraction of debris with planets. Application of this theory has led to the\nrealization that such observations provide a powerful diagnostic that can be\nused not only to refine our understanding of debris disk physics, but also to\nchallenge our understanding of how planetary systems form and evolve.\n