隐式不确定性评估的交互遗传算法提升个性化搜索效率
PDF格式 | 776KB |
更新于2024-08-26
| 29 浏览量 | 举报
本文探讨了在个性化搜索中应用具有隐式不确定性评估的交互式遗传算法（Interactive Genetic Algorithm with Implicit Uncertainty Evaluation, IGA-IUE）的研究。随着交互式进化算法在解决搜索和优化问题上的成功应用，它们在复杂优化场景中的局限性，特别是用户评价疲劳和不确定性问题，成为了关注的焦点。传统方法可能无法有效应对用户在反复参与决策过程时逐渐产生的疲劳感，而隐式评估技术作为一种潜在解决方案，可以减轻这种负担。
作者们，Xiaoyan Sun、Yang Chen、Lin Bao 和 Ruidong Xu，来自中国矿业大学的信息与控制工程学院，他们针对这一挑战提出了创新的方法。IGA-IUE算法通过集成不确定性评估机制，试图在保持用户参与度的同时，处理决策过程中可能出现的偏好不明确或评估结果不稳定的问题。这种方法旨在提高算法的鲁棒性和用户体验，特别是在面对动态变化的个性化需求时。
论文的核心贡献可能包括以下几个方面：
1. **不确定性模型**：构建了一种新的不确定性模型，用于捕捉用户在决策过程中的主观性，并将其转化为算法的适应性因素。
2. **隐式评估**：通过设计算法架构，使得算法能够根据用户的反馈或行为模式进行自我调整，避免显式询问用户的明确偏好，从而减少评价疲劳。
3. **算法优化**：改进了遗传算法的迭代策略，使其能够更好地适应不确定性环境，同时考虑到用户的心理和行为动态。
4. **实验验证**：可能提供了实验数据和案例研究，展示了IUA在个性化搜索任务中的性能提升，以及如何通过隐式评估改善搜索效率和用户满意度。
5. **应用前景**：探讨了该算法在实际场景，如推荐系统、在线学习和复杂决策支持系统中的潜力应用，以及未来可能的研究方向。
这篇研究论文深入研究了交互式遗传算法在处理个性化搜索中的用户不确定性问题，提出了一种创新的隐式评估策略，有望为优化用户体验和提升算法在复杂优化问题上的表现提供有价值的新思路。
Interactive Genetic Algorithm with Implicit
Uncertainty Evaluation for Application in
Personalized Search
Xiaoyan Sun
School of Information and
Control Engineering
China University of Mining and
Technology
Xuzhou, China
xysun78@hotmail.com
Yang
Chen
School of Information and
Control Engineering
China University of Mining and
Technology
Xuzhou, China
fedora.cy@gmail.com
Lin Bao
School of Information and
Control Engineering
China University of Mining and
Technology
Xuzhou, China
baolin_zj@163.com
Ruidong Xu
School of Electrical and Power
Engineering
China University of Mining and
Technology
Xuzhou, China
ruidongxu@163.com
Abstract—Interactive evolutionary algorithms have been used
successfully for search and optimization problems. However,
user evaluation fatigue and uncertainties greatly restrict their
application in complex optimization scenarios. Implicit
evaluation techniques can reduce user fatigue effectively, but
methods for handling evaluation uncertainties have not yet been
explored thoroughly. We study an interactive evolutionary
algorithm with uncertain preferences using an implicit
evaluation model. First, we design a Gaussian function to define
evaluation uncertainties based on interactions. Then, by
combining the evaluation uncertainties and the interactions, we
calculate the preference weights for each variable, with which a
weighted Conditional Preference Net is constructed to provide
quantitative depiction of the user’s evaluation preference. Using
the weighted Net, we estimate the fitness of an individual for the
achievement of the evolutionary process. In the iteration, the Net
is updated according to newly performed interactions to ensure
reliable tracking of the user’s preference. We apply a genetic
algorithm-based interactive evolutionary algorithm to a
personalized search for books. The proposed algorithm
accelerates search speed, alleviates user fatigue, and achieves
higher rates of satisfying solutions.
Keywords—interactive evolutionary algorithm, Conditional
Preference Nets, personalized search, uncertainty
I. INTRODUCTION
Evolutionary algorithms (EAs) are powerful tools for
optimizing complex problems that are difficult for traditional
optimization algorithms to solve [1]. For example, genetic
algorithms (GAs) and particle swarm optimization (PSO) have
proven to be powerful for engineering design problems [2, 3].
For all types of EAs, the design of the fitness function is a
key element. However, for practical problems with qualitative
objectives to be optimized, such as the design of sunglasses or
a house layout, explicitly defined fitness functions are very
difficult, and in some cases even impossible, to obtain [4]. In
such scenarios, conventional EAs face great challenges, and
the designer or the user often evaluates solutions directly
according to their interests or demands. To this end, in 1986,
Dawkins et. al. [
5] proposed a fr
amework for an interactive
evolutionary algorithm (IEA) that imbedded a user’s
evaluations into a EA process.
In interactive evolutionary algorithms (IEAs), traditional
evolutional operators used in EAs are employed according to
the fitness values assigned by a user, rather than calculated by a
mathematical function [6]. IEAs greatly expand the range of
fields in which EAs can be applied by integrating the user’s
evaluations with evolutionary operators; however, evaluation
fatigue and uncertainty on the part of IEA users can result in
fewer generations and/or smaller populations, either of which
would greatly reduce the performance of the IEAs [7].
As surveyed in literature [8], four categories of
uncertainties were considered in evolutionary computation, i.e.
noise, robustness, fitness approximation, and time-varying
fitness functions. Due to the lack of an analytical fitness
function and the dynamic of tracking the evaluation preference
reflected in interactions, uncertainties caused by a mix of the
fitness approximation and the time-varying fitness function are
mainly discussed in our study.
To improve the performance of IEAs, many strategies have
been developed aimed at reducing user fatigue and
strengthening exploration ability [9-11]. For example,
surrogate-based models constructed to approximate user
preferences are the most direct way to reduce the user’s
evaluation burden [12]. Furthermore, to deal with the
evaluation uncertainties of the user, some specific evaluation
modes were proposed by directly adopting intervals or fuzzy
numbers to evaluate the individuals [
13]. These studies,
however, all required the user to assign the fitness explicitly to
each individual during the evolutionary process, so user fatigue
or evaluation boredom was still inevitable.
This work was supported in part by the National Natural Science
Foundation of China under grant 61473298, 61473299 and the Fundamental
Research Funds for the Central Universities under grant 2012QNA58.
XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE
978-1-5386-2726-6/17/$31.00 ©2017 IEEE
下载后可阅读完整内容，剩余7页未读，立即下载
相关推荐
weixin_38599537
- 粉丝: 8
- 资源: 922
我的内容管理
展开
