基于多智能体的分布式推理模型设计与实现

本文探讨了利用多智能体系统(Multi-Agent System,简称MAs)的开放性和协调性建立推理机制模型的思想,研究了协商基础上的推理机制问题.根据协商的思想,设计了分布式推理的算法以及推理机模型,并介绍了网络环境下MAS分布式推理机组件的实现方法.     

作物知识模型构建方法的研究

为了提高作物研究领域专家系统知识的数量与质量，在分析知识模型研究现状的基础上，结合作物的生育特点，提出了作物知识模型建立的基本依据是作物生育时间和农艺技术措施，并依此建立由浅知识模型和深知识模型所组成的油菜三维知识模型，同时对该知识模型的知识表示和模糊推理进行了描述，该知识模型具有一定的通用性。     

基于知识图谱与案例推理的水稻精准施肥推荐模型

目前水稻种植户的施肥行为存在一定的盲目性,会造成肥料浪费和环境污染等问题,对此该研究提出了一种基于知识图谱和案例推理的水稻精准施肥推荐模型,包括推荐定性的施肥方案和定量的施肥量两个阶段。首先,使用PairRE模型获取图谱中全部实体和关系的低维向量表示,并在此基础上依据待种植的水稻品种进行知识推理以得到定性的施肥方案;然后,结合实体向量检索出k个相似案例,通过k个案例进行组合预测,得出具体施肥量数值。由中国知网获取166个环境指标数值明确、施肥过程记录完整的水稻施肥事件用于模型的验证,结果表明,与测试事件的实际施肥方案相违背的部分仅占比10.76%;对于氮肥施用总量、磷肥施用总量、钾肥施用总量和氮肥基叶肥与穗肥运筹比例的预测精度分别达到了92.85%、82.61%、79.17%和90.92%。该施肥推荐模型能够输出详细的施肥方案和精确的施肥量,算法过程可解释性较强,可为水稻精准施肥推荐系统的设计提供支撑。     

现行淤地坝设计洪水计算方法的适用性对比分析

[目的] 洪水是导致淤地坝水毁最根本的原因,深入探讨淤地坝设计中洪水分析计算方法,以期对其发展提供有益参考。[方法] 梳理现行淤地坝设计洪水计算方法,对比分析各种方法的适用特点。对理论性较强的推理公式法在淤地坝设计洪水计算中的适用性进行分析,总结其各类参数的物理意义,提出其运用要点。选取典型算例,研究比较各方法的计算特点及适用性。[结果] 洪水调查法可靠可行但受洪水资料条件限制,经验公式精度和应用具有地区局限性,两种方法均不宜较大范围应用;推理公式法具有一定理论基础和计算精度,且简单实用,具体运用中应注意淤地坝处于特小流域的特点,在计算流域特征值L和J时应判明最远流程中的坡面部分,在计算暴雨递减指数n时应选取由10—60 min的点暴雨n₁值,产汇流参数中μ偏大,m偏小,应在合理范围内相应的倾斜取值。[结论] 推理公式法应当作为淤地坝设计洪水分析计算中的首选方法,需关注淤地坝处于特小流域的特点,把握应用推理公式法的运用要点。     

基于图像规则的玉米病害诊断系统的设计与实现

针对传统病害诊断系统在病症推理过程中借助语言文字描述规则诊断病害让农户难以理解和应用的问题,将图像指认方法与传统文字规则推理技术相结合,建立的病害诊断系统更适合于农户使用,并改善了专家系统的应用效果,利于其在玉米生产中的应用。介绍了基于图像规则的玉米病害诊断系统的设计思想、功能特点和实现方法。     

农业科技服务知识体系的研究与系统构建

由于现有的农业技术服务系统中普遍存在农业知识不完整、数据分析处理能力有限等问题,该文从农业科技服务知识体系的构建入手,针对系统构建的瓶颈问题,结合系统开发的实例对农业可持续发展的体系架构、农业知识的表示方法、多参数的证据不确定性推理等方法进行了探讨．     

Research on Web-GIS Based Intelligent Management Decision-Making System for Soybean

The design of Web-GIS based intelligent management decision-making system for soybean follows the life cycle standard of software engineering. Based on the analysis of the flow of soybean growth technology, the data flow chart was protracted and the function from chart was brought in the course of the designing and realizing the system. By making use of the directed program tool such as VC, Java and multi-media technique the fimctions of decision-making system were realized. It will do a lot of for the theoretical and practical development of intelligent technology of agricultural information of Heilongjiang Province.     

基于信息扩散近似推理的年径流量预测模型

通过分析年径流时间序列的特性,利用信息扩散近似推理描述年径流量间的复杂非线性关系,建立起基于信息扩散近似推理的年径流预测模型.信息扩散近似推理将样本点转换成模糊集,部分弥补了由于数据的不完备性所造成的信息空白,并可以将矛盾模式转换成兼容模式.通过与传统预测方法相比较,发现该模型能够很好地光滑样本数据以及能够较好地发掘知识,有较高的预测精度和推广应用价值.     

Probabilistic risk assessment of agrochemicals in the environment

Concern for the environment has resulted in greater scrutiny of both old and new plant protection products and increased efforts have been directed to developing more rigorous but more realistic procedures for the ecotoxicological risk characterization of these agrochemicals. These techniques include probabilistic analysis of toxicity and exposure data and better understanding of the relationship between structure and function in populations of wildlife and the role of keystone species in maintaining ecosystem functioning. The ecological risk assessment method described here is centered on the use of probabilistic distribution functions that independently describe exposure concentrations and toxicological responses of organisms to the chemical of concern. The distributions are transformed to permit calculation of linear regression parameters. The regression parameters for the two distributions are then used to determine joint probabilities which interrelate the exposure and toxicology data. For ease of presentation the results are presented as an exceedence plot which depicts, based on the exposure data the percent of species likely to be affected and the percent of observations likely to cause this level of effect. In this paper, the use of the method is illustrated using data for chlorpyrifos in North American aquatic environments. These probabilistic risk assessment methods are being assessed for incorporation into assessment procedures in a number of regulatory jurisdictions.     

Inference by exclusion in the red‐tailed black cockatoo (Calyptorhynchus banksii)

Inference by exclusion is the ability to select a given option by excluding the others. When designed appropriately, tests of this ability can reveal choices that cannot be explained by associative processes. Over the past decade, exclusion reasoning has been explored in several non‐human taxonomic groups, including birds, mainly in Corvids and Parrots. To increase our understanding of the taxonomic distribution of exclusion reasoning and, therefore, its evolution, we investigated exclusion performances in red‐tailed black cockatoos (Calyptorhynchus banksii), an Australian relative of the Goffin cockatoo (Cacatua goffini), using a food‐finding task. Cockatoos were required to find a food item hidden in 1 of the 2 experimenter's hands. Following training sessions in which they reliably selected the closed baited hand they had just been shown open, each individual was tested on 4 different conditions. Critical to demonstrating exclusion reasoning was the condition in which they were shown the empty hand and then offered a choice of both closed hands. The performance of all birds was above chance on all experimental conditions but not on an olfactory and/or cuing control condition. The results suggest that the birds might be able to infer by exclusion, although an explanation based on rule learning cannot be excluded. This first experiment in red‐tailed black cockatoo highlights the potential of this species as a model to study avian cognition and paves the pathway for future investigations.