Machine learning techniques produces quite acceptable solutions today. How should be the review evaluation processes of these methods ? How can the methods and results used be easily tested by the reviewers? I have received such criticisms in the evaluation processes of some journals. How easy do you think these methods can be verified ?
I do not think that it is the task of the reviewers to repeat and test the methods and results described in a paper. In most fields of science it is not possible to repeat the experiments or observations of the authors (accelerator results in high-energy physics, chemical technologies, astronomical observations, sociological surveys, ...). The reviewers should check whether scientific methods were applied, whether the research is state-of-the-art and whether the results are presented clearly.
Wolfgang R. Dick
This question actually has two bases:
First, as you mentioned, is it necessary? Second, how can this be made easy for reproducibility studies?
In my opinion; A peer review should be done on the axis of basic equations and literature views. This may not be very possible for some experiments as you mentioned. But in recent years, the topic of reproducibility has been very popular and machine learning model studies can easily turn into a reproducible format. Here are some links that may be of interest to you regarding reproducibility. These processes will also contribute to the referee evaluation.
Thank you for your contribution.
http://rescience.github.io/
https://www.insidehighered.com/quicktakes/2019/09/18/new-journal-focused-reproducibility
Dear Enes Gul
First, you must conduct a thorough literature review of machine learning, its difficulties, and applications in a certain sector. Then you must give something to that. Everything should be written in a word document. The final outcome is a research paper that can be published based on the quality of the research and writing.
Describe the study's objectives and the criteria by which readers should assess the method. Classify the work in terms of a well-known category, such as a formal analysis, a description of a novel learning algorithm, an application of established approaches, or a computer model of human learning.
In addition, have a look at the articles:
1. https://towardsdatascience.com/reviewing-for-machine-learning-conferences-explained-f73bc037babc
2. https://towardsdatascience.com/tips-for-reading-and-writing-an-ml-research-paper-a505863055cf
Kind Regards
Qamar Ul Islam
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking