Text Classification by Combining Different Distance Functions withWeights

Yamada, Takahiro; Yamashita, K.; Ishii, Naohiro; Iwata, Kazunori

doi:10.1109/snpd-sawn.2006.69

Cited by 13 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, in practice, experts in machine learning use different distance functions proposed for general purposes, and they select the best one according to previous results [98]. Also, it is known that learning-based models obtain different classification results when the distance function is changed [99]. The before stated comments are the main reasons for labeling the lazy learning-based models as black-box models.…”

Section: Figurementioning

confidence: 99%

Black-Box vs. White-Box: Understanding Their Advantages and Weaknesses From a Practical Point of View

2019

View full text Add to dashboard Cite

Nowadays, in the international scientific community of machine learning, there exists an enormous discussion about the use of black-box models or explainable models; especially in practical problems. On the one hand, a part of the community defends that black-box models are more accurate than explainable models in some contexts, like image preprocessing. On the other hand, there exist another part of the community alleging that explainable models are better than black-box models because they can obtain comparable results and also they can explain these results in a language close to a human expert by using patterns. In this paper, advantages and weaknesses for each approach are shown; taking into account a stateof-the-art review for both approaches, their practical applications, trends, and future challenges. This paper shows that both approaches are suitable for solving practical problems, but experts in machine learning need to understand the input data, the problem to solve, and the best way for showing the output data before applying a machine learning model. Also, we propose some ideas for fusing both, explainable and blackbox, approaches to provide better solutions to experts in real-world domains. Additionally, we show one way to measure the effectiveness of the applied machine learning model by using expert opinions jointly with statistical methods. Throughout this paper, we show the impact of using explainable and black-box models on the security and medical applications. INDEX TERMS Black-box, white-box, explainable artificial intelligence, deep learning.

show abstract

Section: Figurementioning

confidence: 99%

Black-Box vs. White-Box: Understanding Their Advantages and Weaknesses From a Practical Point of View

2019

View full text Add to dashboard Cite

show abstract

“…As we can see, the idea is not new, although they describe their method as "novel". They mention however that the closest work to theirs is that of [18], which also uses a weighted combination of distances on a text classification task, and where the optimized weights are also computed by applying the genetic algorithms, so we are not sure what the novelty of CSM-GA is.…”

Section: On Using the Genetic Algorithms To Combine Similarity Measurmentioning

confidence: 99%

“…It is important to mention here that in our paper we are not going to discuss the validity of [18], since text classification is not our field of expertise (besides the application details are not presented in [18]), neither will we discuss the similarity measures proposed in that paper. We will not discuss either the correctness of applying the genetic algorithms, or any other optimization technique, to determine the weights of the similarity measures in a combination of similarity measures as a method to enhance the retrieval process in any kind of multimedia search in general.…”

Section: On Using the Genetic Algorithms To Combine Similarity Measurmentioning

confidence: 99%

When Optimization Is Just an Illusion

Fuad

2013

Advanced Data Mining and Applications

View full text Add to dashboard Cite

Bio-inspired optimization algorithms have been successfully applied to solve many problems in engineering, science, and economics. In computer science bio-inspired optimization has different applications in different domains such as software engineering, networks, data mining, and many others. However, some applications may not be appropriate or even correct. In this paper we study this phenomenon through a particular method which applies the genetic algorithms on a time series classification task to set the weights of the similarity measures used in a combination that is used to classify the time series. The weights are supposed to be obtained by applying an optimization process that gives optimal classification accuracy. We show in this work, through examples, discussions, remarks, explanations, and experiments, that the aforementioned method of optimization is not correct and that completely randomly-chosen weights for the similarity measures can give the same classification accuracy.

show abstract