I am tiding up with the below problem, it's a pleasure to have your ideas.
I've written a coding program in two languages, Python and R, but each came to a completely different result. Before jumping to a conclusion, I declare that:
- Every word of the code in two languages has multiple checks and is correct and represents the same thing.
- The used packages in two languages are the same version.
So, what do you think?
The code is about applying deep neural networks for time series data.
Good morning, without the code it is difficilt to know where is the difference I do not use Python i work on R but maybe these difference is due to the stage of spitting dataset do you try to add thr same number in the count of generator of randomly for example seed(1234) (if my memory is good this function is also used in Python language. Were your results and metrics of evaluation totally different? In this case, mayve there is a reliability issue in your model. You should check your data preparation and features selection .
It is possible to have the same code produce different responses in R and Python, especially if the code is using functions or libraries that have different implementations in the two languages.
Some of the reasons why this might occur include:
To avoid these issues, it is important to be aware of the differences between R and Python and to carefully test code in both languages to ensure that it produces the expected results. It may also be helpful to consult the documentation or seek advice from experts in the respective languages.
Throughout the work on "WINDOWS" I have noticed a problem with the floating point of the processors. (which already occurred in the FORTRAN era, in which I also worked)
when I generate random data, and export it for my students to excel, some of them appear as follows
1.0000000000000000000000001
0.99999999999999999999999999999999976
1.0
1.0
and therefore, when the import is performed, it generates a different value.
therefore, for the moment I remain attached to R.
Hello my dear, yes indeed this kind of thing happens and can seem very surprising. For my part, you have to see especially on the side of the subdivision of your database in training and test. Did you split your data using the formula set.seed(0). If not, your training_set and test_set will change every time you run your program and this will permanently change your results. Also, some ML methods are not robust and can give different results between to differents dates for the same parameterization (ex: knn, neuralnetwork)
Find out whether or not R gives the same answer when you rerun your script with a different random seed. (And Python similarly). Are you using the same algorithm and are you using the same parameters?
Dear Richard David Gill , I really appreciate your answer.
Yes, I am using the same algorithms and running under different random seed numbers, resulting in almost similar and close outcomes.
I found out that something is wrong with the TensorFlow library in R and Python since under Keras and PyTorch the results are close.
Thank you very much
- Badges
- Science topic