Data pre-processing function in ANN model
6 vues (au cours des 30 derniers jours)
Afficher commentaires plus anciens
Shing Mei Chiew
le 21 Oct 2022
Modifié(e) : Shing Mei Chiew
le 2 Déc 2022
I need to use ANN model to formulate explicit equations for predicting my data output. I preprocessed my raw data where I normalised my input variables using min-max normalization and logarithmic transformation for my output variable. After that, only I fed the data into the neural network tool in MATLAB. However, after finished training the network, only I realised that the neural network model has built-in preprocessing function which preprocess the input data and post process the output data automatically within the network. Nevertheless, the results from the training and testing is desirable and satisfactory.
Since the ANN has its built-in data preprocessing function, is the first data preprocessing step unneccessary or is it okay to have data preprocessing (manually) before feeding the data into the neural network?
Réponse acceptée
Ben
le 1 Déc 2022
Are you using the Deep Learning Toolbox tools such as DAGNetwork, dlnetwork, trainNetwork and/or custom training loops? In that case the preprocessing is defined by the input layer and can be turned off, for example imageInputLayer(inputSize,Normalization="none")
If that isn't the case could you let us know which network and training functions you are using?
Whether or not it is OK to manually preprocess and let the network also do preprocessing will depend on the particular preprocessing being done. For example if you manually scale the data into the interval [0,1] then it makes no difference if the network training also does this (since the data will already be in [0,1] if you manually preprocessed).
1 commentaire
Plus de réponses (0)
Voir également
Catégories
En savoir plus sur AI for Signals 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!