Learning higher-order logic programs

Machine Learning - A key feature of inductive logic programming is its ability to learn first-order programs, which are intrinsically more expressive than propositional programs. In this paper, we...     

Solving a class of higher-order equations over a group structure

This paper proposes a new symbolic method for solving a class of higher-order equations with an unknown function over the complex domain. Our method exploits the closure property of group structure (for functions) in order to allow an equivalent system of equations to be expressed and solved in the first-order setting.Our work is an initial step towards the relatively unexplored realm of higher-order constraint solving, in general; and higher-order equational solving, in particular. We shall provide some theoretical background for the proposed method, and also prototype an implementation under Mathematica.     

Learning associations between natural groups of input and output with a neurofuzzy structure

A new modeling approach that finds the associations between natural groups of input and output is proposed. In the new method, input and output are clustered separately by means of Fuzzy C-Means (FCM) algorithm. Then, the learning algorithm identifies the fuzzy rules by relating the resulting fuzzy sets in input and output spaces by using a neurofuzzy architecture. A modified version of classical simulated annealing algorithm is used in order to identify the relative weights of system input variables. The proposed approach is applied to a highly nonlinear function and successful result is achieved.     

Learning from natural instructions

Machine learning is traditionally formalized and investigated as the study of learning concepts and decision functions from labeled examples, requiring a representation that encodes information about the domain of the decision function to be learned. We are interested in providing a way for a human teacher to interact with an automated learner using natural instructions, thus allowing the teacher to communicate the relevant domain expertise to the learner without necessarily knowing anything about the internal representations used in the learning process. In this paper we suggest to view the process of learning a decision function as a natural language lesson interpretation problem, as opposed to learning from labeled examples. This view of machine learning is motivated by human learning processes, in which the learner is given a lesson describing the target concept directly and a few instances exemplifying it. We introduce a learning algorithm for the lesson interpretation problem that receives feedback from its performance on the final task, while learning jointly (1) how to interpret the lesson and (2) how to use this interpretation to do well on the final task. traditional machine learning by focusing on supplying the learner only with information that can be provided by a task expert. We evaluate our approach by applying it to the rules of the solitaire card game. We show that our learning approach can eventually use natural language instructions to learn the target concept and play the game legally. Furthermore, we show that the learned semantic interpreter also generalizes to previously unseen instructions.     

Learning a representation with the block-diagonal structure for pattern classification

Pattern Analysis and Applications - Sparse-representation-based classification (SRC) has been widely studied and developed for various practical signal classification applications. However, the...     

Raising natural frequencies of a structure via surface-grooving technique

Structural dynamics modification (SDM) is a very effective technique to improve a structure’s dynamic characteristics by adding or removing auxiliary structures, changing material properties and shape of structure. Among the SDM techniques, changing or modifying structure shape to raise its natural frequencies has been mostly relied on engineer’s experience and time-consuming trial-and-error process. To develop a systematic method to modify structure shape, surface-grooving technique is studied. In this work, the shape of base structure is modified to improve its dynamic characteristics such as natural frequencies via surface-grooving technique. Grooving shape is formed by merging the neighboring small embossed elements after analyzing frequency increment sensitivities of all the surrounding embossed elements. All this process is targeted to pack in a software to get an optimum grooving shape automatically. In this package, the initial grooving position starts from the element having the highest modal strain energy then it expands into neighboring elements. The range of grooving area for checking its frequency sensitivities is restricted only to their surrounding elements to reduce its computation effort. The developed algorithm was tested with an L-shaped plate and hard disk drive (HDD) cover to raise its natural frequency by giving some groove on its surface. Also, the grooved HDD cover design was manufactured using rapid prototyping and tested to prove the effectiveness of the surface grooving as a SDM tool.     

Learning the nonlinearity of neurons from natural visual stimuli

Learning in neural networks is usually applied to parameters related to linear kernels and keeps the nonlinearity of the model fixed. Thus, for successful models, properties and parameters of the nonlinearity have to be specified using a priori knowledge, which often is missing. Here, we investigate adapting the nonlinearity simultaneously with the linear kernel. We use natural visual stimuli for training a simple model of the visual system. Many of the neurons converge to an energy detector matching existing models of complex cells. The overall distribution of the parameter describing the nonlinearity well matches recent physiological results. Controls with randomly shuffled natural stimuli and pink noise demonstrate that the match of simulation and experimental results depends on the higher-order statistical properties of natural stimuli.     

Towards a theory of bisimulation for the higher-order process calculi

In this paper, a labelled transition semantics for higher-order process calculi is studied. The labelled transition semantics is relatively clean and simple, and corresponding bisimulation equivalence can be easily formulated based on it. And the congruence properties of the bisimulation equivalence can be proved easily. To show the correspondence between the proposed semantics and the well-established ones, the bisimulation is characterized as a version of barbed equivalence and a version of context bisimulation.     

Discrete modelling of uncertain continuous systems having an interval structure using higher-order integrators

In this paper, a higher-order integrator approach is proposed to obtain an approximate discrete-time transfer function for uncertain continuous systems having interval uncertainties. Because of the simple algebraic operations of this approach, the resulting discrete model is a rational function of the uncertain parameters. The problem of non-linearly coupled coefficients with exponential nature occurring in the exact discretetime transfer function is therefore circumvented. Furthermore, the interval structure of the uncertain continuous-time system is preserved in the resulting discrete model by using this approach. Formulae to obtain the lower and upper bounds for the coefficients of the discrete interval system are derived, so that digital simulation and design for the uncertain continuous systems can be performed by using the available robustness results in the discrete-time domain.     

Evolutionary optimization of maximizing the difference between two natural frequencies of a vibrating structure

This paper deals with the evolutionary optimization of maximizing the difference between two natural frequencies of a vibrating structure. Two new criteria, namely the material efficiency criterion and the smooth change criterion, are derived for solving this kind of evolutionary optimization problem. Using these two new criteria, the evolutionary optimization method has been further extended and applied to maximize the difference between the fundamental and the second natural frequencies of a structure under both plane stress and thin plate flexural bending conditions. The related results demonstrated that the extended evolutionary structural optimization method is useful in and applicable to dealing with the evolutionary optimization of maximizing the difference between two natural frequencies of a vibrating structure. Moreover, the results also indicated that owing to the different mechanism between plane stress and thin plate flexural bending conditions, the optimal topologies, the normalized difference between two natural frequencies and the normalized material efficiency are different for a vibrating structure under these two different situations.     

用于因果分析的混合贝叶斯网络结构学习

目前主要结合扩展的熵离散化方法和打分一搜索方法进行混合贝叶斯网络结构学习，算法效率和可靠性低，而且易于陷入局部最优结构。针对问题建立了一种新的混合贝叶斯网络结构迭代学习方法．在迭代中，基于父结点结构和Gibbs sampling进行混合数据聚类，实现对连续变量的离散化，再结合贝叶斯网络结构优化调整，使贝叶斯网络结构序列逐渐趋于稳定，可避免使用扩展的熵离散化和打分——搜索所带来的主要问题．     

基于Abaqus的水下结构声辐射仿真方法

提出一种在Abaqus中直接完成辐射声场仿真的方法,该方法可大大减少流场网格层数.利用无限元的计算结果求解任意远处球形声场的声压;结合弹性球壳的解析解算例验证该方法的正确性,并将该方法拓展到细长型结构中.结果表明：基于Abaqus无限元的水下结构声辐射的计算方法简单、可行;采用有限元与无限元相结合的方法可以大幅度缩小流场区域.     

Typing termination in a higher-order concurrent imperative language

We propose means to predict termination in a higher-order imperative and concurrent language à la ML. We follow and adapt the classical method for proving termination in typed formalisms, namely the realizability technique. There is a specific difficulty with higher-order state, which is that one cannot define a realizability interpretation simply by induction on types, because applying a function may have side-effects at types not smaller than the type of the function. Moreover, such higher-order side-effects may give rise to computations that diverge without resorting to explicit recursion. We overcome these difficulties by introducing a type and effect system for our language that enforces a stratification of the memory. The stratification prevents the circularities in the memory that may cause divergence, and allows us to define a realizability interpretation of the types and effects, which we then use to establish that typable sequential programs in our system are guaranteed to terminate, unless they use explicit recursion in a divergent way. We actually prove a more general fairness property, that is, any typable thread yields the scheduler after some finite computation. Our realizability interpretation also copes with dynamic thread creation.     

Learning about natural human–robot interaction styles

If we are to achieve natural human–robot interaction, we may need to complement current vision and speech interfaces. Touch may provide us with an extra tool in this quest. In this paper we demonstrate the role of touch in interaction between a robot and a human. We show how infrared sensors located on robots can be easily used to detect and distinguish human interaction, in this case interaction with individual children. This application of infrared sensors potentially has many uses; for example, in entertainment or service robotics. This system could also benefit therapy or rehabilitation, where the observation and recording of movement and interaction is important. In the long term, this technique might enable robots to adapt to individuals or individual types of user.     

Intention structure and extended responses in a portable natural language interface

This paper describes discourse processing inKing Kong, a portable natural language interface.King Kong enables users to pose questions and issue commands to a back end system. The notion of a discourse is central toKing Kong, and underlies much of the intelligent assistance thatkong provides to its users.kong's approach to modeling discourse is based on the work of Grosz and Sidner (1986). We extend Grosz and Sidner's framework in several ways, principally to allow multiple independent discourse contexts to remain active at the same time. This paper also describesKing Kong's method of intention recognition, which is similar to that described in Kautz and Allen (1986) and Carberry (1988). We demonstrate that a relatively simple intention recognition component can be exploited by many other discourserelated mechanisms, for example to disambiguate input and resolve anaphora. In particular, this paper describes in detail the mechanism inKing Kong that uses information from the discourse model to form a range of cooperative extended responses to queries in an effort to aid the user in accomplishing her goals.Judith Schaffer Sider received her Bachelor of Arts degree in Computer Science and Linguistics and Cognitive Science from Brandeis University. Since 1987 she has been a member of the technical staff at the MITRE Corporation, where she works on King Kong, the natural language interface under development there. The joint research with John D. Burger described in this volume reflects some of her work in the areas of cooperative responding and plan recognition.John D. Burger is a Project Leader at the MITRE Corporation and an instructor at Boston University. He received a Bachelor of Science degree in Mathematics and Computer Science from Carnegie Melon University. His research interests lie in the fields of natural language processing and intelligent multimedia interfaces. The joint work with Judith Schaffer Sider described in this volume reflects his interest in making use of discourse models in practical intelligent interfaces.     

Extending a brainiac prover to lambda-free higher-order logic

Decades of work have gone into developing efficient proof calculi, data structures, algorithms, and heuristics for first-order automatic theorem proving. Higher-order provers lag behind in terms of efficiency. Instead of developing a new higher-order prover from the ground up, we propose to start with the state-of-the-art superposition prover E and gradually enrich it with higher-order features. We explain how to extend the prover’s data structures, algorithms, and heuristics to \(\lambda \)-free higher-order logic, a formalism that supports partial application and applied variables. Our extension outperforms the traditional encoding and appears promising as a stepping stone toward full higher-order logic.

     

On the capabilities of higher-order neurons: a radial basis function approach

Higher-order neurons with k monomials in n variables are shown to have Vapnik-Chervonenkis (VC) dimension at least nk + 1. This result supersedes the previously known lower bound obtained via k-term monotone disjunctive normal form (DNF) formulas. Moreover, it implies that the VC dimension of higher-order neurons with k monomials is strictly larger than the VC dimension of k-term monotone DNF. The result is achieved by introducing an exponential approach that employs gaussian radial basis function neural networks for obtaining classifications of points in terms of higher-order neurons.