Root mean square error (RMSE) is often used in ANFIS. It often refers to the training and checking error. How about the average testing error? For instance, in Matlab anfisedit GUI, it normally shows the average testing error after clicking "testing FIS".
Can average testing error be used to validate the FIS model?
Thanks!
Dear Zhe Zhang,
Please see Answer of similar Question posed trough ResearchGate, here is link:
How can average testing error for checking data-sets be...
https://www.researchgate.net/.../How_can_average_testing_error_f...
Also i suggest you to see attached files and publications in topics.
Best regards
Article Integrated Models for Information Communication Systems and ...
Article Adaptive Neuro-Fuzzy Inference System in Predicting the Succ...
Article Laboratory Biogas Production from Kitchen Wastes and Applyin...
Conference Paper Adaptive neuro-fuzzy Inference system for predicting Norovir...
Article Adaptive neuro-fuzzy inference system for predicting alpha b...
Thank you very much for the articles!
I read the articles, and most of them used the training error( RMSE) to validate the ANFIS model, e.g. plotting every epoch training error. One article mentioned about the average testing error.
"After FIS training, validate the model using a testing or checking data that differs from the one you used to train the FIS. Average testing error of training and testing data in the ANFIS model 1 are 0.86 and 0.8, respectively that have been shown in FIG.15 and FIG.16, respectively. "
According to the text above, I assume that we can use testing or checking data to validate the model, right? For instance, I only used checking data. However, the author didn't explain the meaning of average testing error. I would assume that is the RMSE between FIS training results and the checking data.
Thank you very much!
ARMSE is the means RMSE obtained at each iteration.
You can't use it to validate your model because it gives information on the speed of convergence towards the optimal solution.
For example if the error decreases rapidly from a large value to a small ARMSE value will be larger otherwise it will be in the vicinity of RMSE
Is average testing error of anfis controller is considered as RMSE? Pls tell me
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling