Evolving multi-user fuzzy classifier systems integrating human uncertainty and expert knowledge
نویسندگان
چکیده
Evolving classifiers and especially evolving fuzzy have been established as a prominent technique for addressing the recent demands in building an incremental open-loop manner, e.g. purpose of processing data streams online. So far, focus lies on which are obtained based input and/or feedback form target labels provided by single user/expert. In this paper, we propose three variants multi-user classifier systems (EFCS-MU), where multiple users may provide their label feedback: i) ensembled single-user system, allows separate training per user embeds advanced aggregation strategy (→ ensembling model level), ii) consensus all-user joint is all labelings iii) shift-work classical concept. The incrementally evolved single-pass learning approach embedding autonomous evolution new rules demand; it integrates unsupervised clustering rule partitioning, thus same partition classifiers, only consequents class confidence vectors typically differ among due to different labelings. This offers direct explainability varying users’ annotation behaviors. possible experience levels relation process behind ambiguities handled proper integration uncertainty into update classifier(s). Furthermore, concept presented how adequately integrate possibly available expert particular newly (on-the-fly) arising (or advance several classes). Finally, on-line active (oAL) demonstrated, select most important samples be labelled reduce labeling costs, ensuring economically practicable applicability. was successfully evaluated two real-world application scenarios, one stemming visual inspection scenario, four check quality imprint compact discs affected levels, from nursery school employment ranking application, introduced later. results insights performance behavior under circumstances (with without rules, integration, labelling budgets etc.), including comparisons accuracy trends versus economy effort.
منابع مشابه
investigation of single-user and multi-user detection methods in mc-cdma systems and comparison of their performances
در این پایان نامه به بررسی روش های آشکارسازی در سیستم های mc-cdma می پردازیم. با توجه به ماهیت آشکارسازی در این سیستم ها، تکنیک های آشکارسازی را می توان به دو دسته ی اصلی تقسیم نمود: آشکارسازی سیگنال ارسالی یک کاربر مطلوب بدون در نظر گرفتن اطلاعاتی در مورد سایر کاربران تداخل کننده که از آن ها به عنوان آشکارساز های تک کاربره یاد می شود و همچنین آشکارسازی سیگنال ارسالی همه ی کاربران فعال موجود در...
Evolving Ensemble Fuzzy Classifier
The concept of ensemble learning offers a promising avenue in learning from data streams under complex environments because it addresses the bias and variance dilemma better than its single model counterpart and features a reconfigurable structure, which is well suited to the given context. While various extensions of ensemble learning for mining non-stationary data streams can be found in the ...
متن کاملMedical Knowledge, Fuzzy Sets, and Expert Systems
The term medical knowledge is a superimposed concept for the relationships between symptoms and diagnoses a physician may find in books, journals, monographs, but also in practical experience. In the second half of the 20th century medical knowledge was also stored in computer systems. To assist physicians in medical decision-making and attendance medical expert systems have been constructed th...
متن کاملKnowledge acquisition for fuzzy expert systems
Expert systems have been successfully applied to a wide variety of application domains. To achieve better performance, researchers have tried to employ fuzzy logic to the development of expert systems. However, as fuzzy rules and membership functions are difficult to define, most of the existing tools and environments for expert systems do not support fuzzy representation and reasoning. Thus, i...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Information Sciences
سال: 2022
ISSN: ['0020-0255', '1872-6291']
DOI: https://doi.org/10.1016/j.ins.2022.03.014