Any relative information concerning analysis for small data (except for linear regression,clustering & principal component analysis is appreciated )
Sorry,i thought it was clear enough.i have a data that i want to analyze using statistical method.& because i already know (from previous studies) that there is a strong relations between my variables & my response variable,I used neural networks for prediction .but after reading some publications i start questioning my self,is it right to use ANN for small data or it's just for big Data.
just so you know R² was about 90% in both training & validation.but maybe because it's a small !
How much is the sample size. Mostly, for small sample size, recommended tool is Smart PLS
Regards
Dear Asmae,
ANNs are useful when:
- There is no alternative direct (batch) optimization method (such as multilinear regression),
- The data model optimization problem is nonlinear.
Because ANNs often require iterative optimization methods that are time-consuming and may be sub-optimal (local minima).
Nevertheless, neural networks can address small datasets. But take care not to overfit your data in any case! To this end, prefer incremental learning methods. Follow:
- https://www.researchgate.net/post/What_is_online_training_in_convolutional_neural_networks
- Lütz et al., "I Want To Know More — Efficient Multi-Class Incremental Learning Using Gaussian Processes", 2013 - http://hera.inf-cv.uni-jena.de:6680/pdf/Luetz13:IWT.pdf
Best Regards
- Badges
- Science topic