Heuristic Learning Based on Genetic Programming

In this paper we present an approach to learning heuristics based on Genetic Programming (GP) which can be applied to problems in the VLSI CAD area. GP is used to develop a heuristic that is applied to the problem instance instead of directly solving the problem by application of GP. The GP-based heuristic learning method is applied to one concrete field from the area of VLSI CAD, i.e. minimization of Binary Decision Diagrams (BDDs). Experimental results are given in order to demonstrate that the GP-based method leads to high quality results that outperform previous methods while the run-times of the resulting heuristics do not increase. Furthermore, we show that by clever adjustment of parameters, further improvements such as the saving of about 50% of the run-time for the learning phase can be achieved.     

基于BOA-ELM的混凝土抗压强度预测研究

为控制控制混凝土生产成本,在混凝土拌和期限制抗压强度不足的缺陷构建产出,可以有效降低原料的浪费,是节能降耗的关键方法之一。针对混凝土抗压强度的传统测量方法严重滞后的问题,提出了基于贝叶斯优化极限学习机(BOA-ELM)的混凝土抗压强度预测方法。首先,分析了混凝土拌和过程中对抗压强度预测值实时获得的需求。以各物料的用量为分析基础,28天标准养护后混凝土抗压强度值为预测目标,设计了基于极限学习机的强度预测模型。其次,为进一步提高模型的稳定性以及准确行,提出基于贝叶斯优化的极限学习机模型,根据模型超参数的分布特征,以高斯过程作为超参的先验分布,预测误差最小化作为目标,寻找最优的模型超参。最后,在实际施工产生的C50标号混凝土数据集上测试文中模型,并对比分析了其他预测模型和寻优算法。结果表明,结合了贝叶斯优化的极限学习机预测模型相较于经典算法具有更高的预测准确性和模型训练的高效性。     

基于金枪鱼群算法优化极限学习机的混凝土抗压强度预测

混凝土抗压强度是建筑结构设计与评价一个重要指标,它直接关乎建筑的质量与安全。为解决现有机器学习模型对其预测存在预测耗时长、精度不够高,不能很好地满足施工现场对混凝土抗压强度预测实时性与准确性要求的问题,提出一套基于新式仿生算法金枪鱼群算法优化极限学习机（TSO-ELM）的混凝土抗压强度预测方法。该方法通过对ELM隐藏层初始参数中的连接权值与偏置值使用TSO进行寻优,有效提升了ELM的预测准确度。在仿真实验部分,通过两组混凝土数据集对ELM的预测速度、TSO的寻优能力、TSO-ELM模型的泛化性逐一进行验证。结果表明,该方法可以有效提高预测的速度与精准度,迭代次数更少,同时具有良好的泛化性,为现场施工及时进行混凝土抗压强度的预测提供了一种新方法。     

Hybrid intelligent systems for predicting software reliability

In this paper, we propose novel recurrent architectures for Genetic Programming (GP) and Group Method of Data Handling (GMDH) to predict software reliability. The effectiveness of the models is compared with that of well-known machine learning techniques viz. Multiple Linear Regression (MLR), Multivariate Adaptive Regression Splines (MARS), Backpropagation Neural Network (BPNN), Counter Propagation Neural Network (CPNN), Dynamic Evolving Neuro-Fuzzy Inference System (DENFIS), TreeNet, GMDH and GP on three datasets taken from literature. Further, we extended our research by developing GP and GMDH based ensemble models to predict software reliability. In the ensemble models, we considered GP and GMDH as constituent models and chose GP, GMDH, BPNN and Average as arbitrators. The results obtained from our experiments indicate that the new recurrent architecture for GP and the ensemble based on GP outperformed all other techniques.     

基于变精度粗糙集的煤与瓦斯突出预测

煤与瓦斯突出是受诸多因素影响的复杂问题,为了提高其预测的准确性,本文以变精度粗糙集理论中的决策表为主要工具,首先将影响煤与瓦斯突出的检测信号作为预测的条件属性集,煤与瓦斯突出量作为对预测的决策属性,建立决策表,然后利用小生境遗传算法适合于进行多峰值函数优化的特点,提出了一种基于小生境遗传算法的粗糙集属性约简方法,用于求解决策表的多个约简,进而进行值约简后抽取出预测规则。算例结果说明了本算法的正确性和可行性。     

Prediction of relative position of CT slices using a computational intelligence system

One of the most common techniques in radiology is the computerized tomography (CT) scan. Automatically determining the relative position of a single CT slice within the human body can be very useful. It can allow for an efficient retrieval of slices from the same body region taken in other volume scans and provide useful information to the non-expert user. This work addresses the problem of determining which portion of the body is shown by a stack of axial CT image slices. To tackle this problem, this work proposes a computational intelligence system that combines semantics-based operators for Genetic Programming with a local search algorithm, coupling the exploration ability of the former with the exploitation ability of the latter. This allows the search process to quickly converge towards (near-)optimal solutions. Experimental results, using a large database of CT images, have confirmed the suitability of the proposed system for the prediction of the relative position of a CT slice. In particular, the new method achieves a median localization error of 3.4 cm on unseen data, outperforming standard Genetic Programming and other techniques that have been applied to the same dataset. In summary, this paper makes two contributions: (i) in the radiology domain, the proposed system outperforms current state-of-the-art techniques; (ii) from the computational intelligence perspective, the results show that including a local searcher in Geometric Semantic Genetic Programming can speed up convergence without degrading test performance.     

Evolutionary Synthesis of Logic Circuits Using Information Theory

In this paper, we propose the use of Information Theory as thebasis for designing a fitness function for Boolean circuit designusing Genetic Programming. Boolean functions are implemented byreplicating binary multiplexers. Entropy-based measures, such asMutual Information and Normalized Mutual Information areinvestigated as tools for similarity measures between the targetand evolving circuit. Three fitness functions are built over aprimitive one. We show that the landscape of Normalized MutualInformation is more amenable for being used as a fitness functionthan simple Mutual Information. The evolutionary synthesizedcircuits are compared to the known optimum size. A discussion ofthe potential of the Information-Theoretical approach is given.     

A genetic type-2 fuzzy logic based system for the generation of summarised linguistic predictive models for financial applications

The global economic meltdown of the late 2000s exposed many organisations around the world, this drove the need to build robust frameworks for predicting and assessing risks in financial applications. Such predictive frameworks helped organisations to increase the quality and quantity of their transactions hence increasing the revenues and reducing the risks. Many organisations around the World still use statistical regression techniques which are well established for many problems such as fraud detection or risk analysis. However, recent years have seen the application of computational intelligence techniques to develop predictive models for financial applications. Some of the computational intelligence techniques like neural networks provide good predictive models, nevertheless they are considered as black box models which do not provide an easy to understand reasoning about a given decision or even a summary of the generated model. However, in the current economic situation, transparency became an important factor where there is a need to fully understand and analyze a given financial model. In this paper, we will present a Genetic Type-2 Fuzzy Logic System (FLS) for the modeling and prediction of financial applications. The proposed system is capable of generating summarized models from a specified number of linguistic rules, which enables the user to understand the generated financial model. The system is able to use this summarized model for prediction within financial applications. We have performed several evaluations in two distinctive financial domains, one for the prediction of good/bad customers in a financial real-world lending application and the other domain was in the prediction of arbitrage opportunities in the stock markets. The proposed Genetic type-2 FLS has outperformed white box models like the Evolving Decision Rule procedure (which is a white based on Genetic Programming and decision trees) and gave a comparable performance to black box models like neural networks while the proposed genetic type-2 FLS provided a white box model which is easy to understand and analyse by the lay user.     

Learning to solve planning problems efficiently by means of genetic programming.

Declarative problem solving, such as planning, poses interesting challenges for Genetic Programming (GP). There have been recent attempts to apply GP to planning that fit two approaches: (a) using GP to search in plan space or (b) to evolve a planner. In this article, we propose to evolve only the heuristics to make a particular planner more efficient. This approach is more feasible than (b) because it does not have to build a planner from scratch but can take advantage of already existing planning systems. It is also more efficient than (a) because once the heuristics have been evolved, they can be used to solve a whole class of different planning problems in a planning domain, instead of running GP for every new planning problem. Empirical results show that our approach (EvoCK) is able to evolve heuristics in two planning domains (the blocks world and the logistics domain) that improve PRODIGY4.0 performance. Additionally, we experiment with a new genetic operator --Instance-Based Crossover--that is able to use traces of the base planner as raw genetic material to be injected into the evolving population.     

Prediction and multi-objective optimization of high-strength concrete parameters via soft computing approaches

The optimization of composite materials such as concrete deals with the problem of selecting the values of several variables which determine composition, compressive stress, workability and cost etc. This study presents multi-objective optimization (MOO) of high-strength concretes (HSCs). One of the main problems in the optimization of HSCs is to obtain mathematical equations that represents concrete characteristic in terms of its constitutions. In order to solve this problem, a two step approach is used in this study. In the first step, the prediction of HSCs parameters is performed by using regression analysis, neural networks and Gen Expression Programming (GEP). The output of the first step is the equations that can be used to predict HSCs properties (i.e. compressive stress, cost and workability). In order to derive these equations the data set which contains many different mix proportions of HSCs is gathered from the literature. In the second step, a MOO model is developed by making use of the equations developed in the first step. The resulting MOO model is solved by using a Genetic Algorithm (GA). GA employs weighted and hierarchical method in order to handle multiple objectives. The performances of the prediction and optimization methods are also compared in the paper.     

Severe damage recovery in evolving soft robots through differentiable programming

Genetic Programming and Evolvable Machines - Biological systems are very robust to morphological damage, but artificial systems (robots) are currently not. In this paper we present a system based...     

Two-Tier genetic programming: towards raw pixel-based image classification

Classifying images is of great importance in machine vision and image analysis applications such as object recognition and face detection. Conventional methods build classifiers based on certain types of image features instead of raw pixels because the dimensionality of raw inputs is often too large. Determining an optimal set of features for a particular task is usually the focus of conventional image classification methods. In this study we propose a Genetic Programming (GP) method by which raw images can be directly fed as the classification inputs. It is named as Two-Tier GP as every classifier evolved by it has two tiers, the other for computing features based on raw pixel input, one for making decisions. Relevant features are expected to be self-constructed by GP along the evolutionary process. This method is compared with feature based image classification by GP and another GP method which also aims to automatically extract image features. Four different classification tasks are used in the comparison, and the results show that the highest accuracies are achieved by Two-Tier GP. Further analysis on the evolved solutions reveals that there are genuine features formulated by the evolved solutions which can classify target images accurately.     

Discovering Multiple Diagnostic Rules from Coronary Heart Disease Database using Automatically Defined Groups

Much of the research on extracting rules from a large amount of data has focused on the extraction of a general rule that covers as many data as possible. In the field of health care, where people’s lives are at stake, it is necessary to diagnose appropriately without overlooking the small number of patients who show different symptoms. Thus, the exceptional rules for rare cases are also important. From such a viewpoint, multiple rules, each of which covers a part of the data, are needed for covering all data. In this paper, we describe the extraction of such multiple rules, each of which is expressed by a tree structural program. We consider a multi-agent approach to be effective for this purpose. Each agent has a rule that covers a part of the data set, and multiple rules which cover all data are extracted by multi-agent cooperation. In order to realize this approach, we propose a new method for rule extraction using Automatically Defined Groups (ADG). The ADG, which is based on Genetic Programming, is an evolutionary optimization method of multi-agent systems. By using this method, we can acquire both the number of necessary rules and the tree structural programs which represent these respective rules. We applied this method to a database used in the machine learning field and showed its effectiveness. Moreover, we applied this method to medical data and developed a diagnostic system for coronary heart diseases     

Genetic Programming with a Genetic Algorithm for Feature Construction and Selection

The use of machine learning techniques to automatically analyse data for information is becoming increasingly widespread. In this paper we primarily examine the use of Genetic Programming and a Genetic Algorithm to pre-process data before it is classified using the C4.5 decision tree learning algorithm. Genetic Programming is used to construct new features from those available in the data, a potentially significant process for data mining since it gives consideration to hidden relationships between features. A Genetic Algorithm is used to determine which such features are the most predictive. Using ten well-known datasets we show that our approach, in comparison to C4.5 alone, provides marked improvement in a number of cases. We then examine its use with other well-known machine learning techniques.     

Evolution of Gaussian Process kernels for machine translation post-editing effort estimation

In many Natural Language Processing problems the combination of machine learning and optimization techniques is essential. One of these problems is the estimation of the human effort needed to improve a text that has been translated using a machine translation method. Recent advances in this area have shown that Gaussian Processes can be effective in post-editing effort prediction. However, Gaussian Processes require a kernel function to be defined, the choice of which highly influences the quality of the prediction. On the other hand, the extraction of features from the text can be very labor-intensive, although recent advances in sentence embedding have shown that this process can be automated. In this paper, we use a Genetic Programming algorithm to evolve kernels for Gaussian Processes to predict post-editing effort based on sentence embeddings. We show that the combination of evolutionary optimization and Gaussian Processes removes the need for a-priori specification of the kernel choice, and, by using a multi-objective variant of the Genetic Programming approach, kernels that are suitable for predicting several metrics can be learned. We also investigate the effect that the choice of the sentence embedding method has on the kernel learning process.

     

Evolving Recursive Programs by Using Adaptive Grammar Based Genetic Programming

Genetic programming (GP) extends traditional genetic algorithms to automatically induce computer programs. GP has been applied in a wide range of applications such as software re-engineering, electrical circuits synthesis, knowledge engineering, and data mining. One of the most important and challenging research areas in GP is the investigation of ways to successfully evolve recursive programs. A recursive program is one that calls itself either directly or indirectly through other programs. Because recursions lead to compact and general programs and provide a mechanism for reusing program code, they facilitate GP to solve larger and more complicated problems. Nevertheless, it is commonly agreed that the recursive program learning problem is very difficult for GP. In this paper, we propose techniques to tackle the difficulties in learning recursive programs. The techniques are incorporated into an adaptive Grammar Based Genetic Programming system (adaptive GBGP). A number of experiments have been performed to demonstrate that the system improves the effectiveness and efficiency in evolving recursive programs. Communicated by: William B. Langdon An erratum to this article is available at .     

An evolutionary approach to rank class association rules with feedback mechanism

In this paper, we propose an evolutionary associative classification method by considering both adjustment of the order of the whole set of rules and refinement of the power of each single rule. We discover an interesting phenomenon that the classification performance could be improved if we import some prior-knowledge to re-rank the association rules, where the prior-knowledge could be some equations generated by combing the support and confidence values with various functions. We make use of Genetic Network Programming to automatically search the equation space for prior-knowledge. In addition to rank the rules by equations globally, we also develop a feedback mechanism to adjust the rules locally, by giving some rewards to good rules and penalties to bad ones. Because the proposed method is based on evolutionary computation, we could gradually refine the power of each rule so that it could affect the classification results more precisely. The experimental results on UCI benchmark datasets show that the proposed method could improve the classification accuracies effectively.     

基于GNP的多Agent人工股市模型

结合奥地利学派的经济思想，本文介绍了一种新的基于GNP算法的多Agent人工股市模型。该模型采用GNP算法来模拟交易个体的行为模式，进化他们的决策规则；同时在设计上强化Agent的异质性，并利用GA算法来优化模型参数。仿真结果表明，GNP-ASM模型表现出很好的统计性能，能够体现真实股市的一些基本特征。     

基于速度预测的异构网络切换算法

为高速移动的用户提供一种基于速度预测的异构网络垂直切换算法,使用场景如公路的直线路段.根据直线道路场景的特点,结合驻留时间算法设计了一种基于速度预测的垂直切换算法.该算法通过创建速度矩阵以及与速度矩阵相对应的权重矩阵,经过数学运算,求得下一时刻的速度和终端的位置,然后结合传统算法得到切换判决.由仿真得到,在相同的环境及可靠的信号强度下,与经典的算法相比较,新算法虽然增加了切换次数,但有效减少了切换延时.