How can I turn off a "self learning" function of Matlab?
4 vues (au cours des 30 derniers jours)
Afficher commentaires plus anciens
DenisK
le 11 Mai 2018
Réponse apportée : DenisK
le 11 Mai 2018
Hello,
I've got an assignment where I was asked to compare time required for different convolutions methods. Surprisingly, time of linear convolution was significantly shorter rather than overlap add method. I heard that matlab has kind of "self learning" mechanism, that, I assume, interfered to the whole process and insert kind of unwanted correction that changed time that takes to algorithm to run. Can I turn off this mechanism, if such actually exists?
Thank you in advance, Denis
0 commentaires
Réponse acceptée
Walter Roberson
le 11 Mai 2018
I suspect you might be referring to the Just In Time Engine (JIT). There is an undocumented
feature('jit', 'off')
Be sure to "clear all" after that in order to remove everything from MATLAB's cache.
However, if you are using data files, this will not remove the data files from your operating system's cache -- for reasonable sized files, your operating system probably left the file hanging around in memory after it was read, under the theory that it might be needed again afterwards.
You should probably be looking at the timeit() function to try to isolate how long something "really" takes: it runs the code several times to try to account for cache and JIT effects.
0 commentaires
Plus de réponses (2)
DenisK
le 11 Mai 2018
1 commentaire
Walter Roberson
le 11 Mai 2018
Nothing like that sounds familiar to me, especially not for any convolution methods. Perhaps something to do with "deep learning"??
Voir également
Catégories
En savoir plus sur Logical dans Help Center et File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!