How to create target data for neural netwrok traing
1 visualización (últimos 30 días)
Mostrar comentarios más antiguos
roma marfatia
el 15 de Abr. de 2014
Comentada: Greg Heath
el 23 de Abr. de 2014
i have training data, 8 samples, where each sample contains 32 features, and total number of class in which they should fall are 8 classes. how should i create my input and target data
0 comentarios
Respuesta aceptada
Greg Heath
el 16 de Abr. de 2014
1. Nomenclature: You have one sample. The sample contains N=8, I=32-dimensional input examples from c=8 classes.
2. Are they all from different classes or are there empty classes and classes with multiple members?
3. 8 examples define, at most, a 7-dimensional input space. Therefore, if you cannot get more training examples, you should reduce the dimensionality to, at most, 7.
4. The input and target matrices should have the sizes
[ I N ] = size(input)
[ c N ] = size(target)
where the target columns are columns of the unit matrix eye(c).
5. Ideally you would like N >~ 15*I. So find more data. Otherwise use a simple classifier like linear or nearest neighbor.
Hope this helps.
Thank you for formally accepting my answer
Greg
2 comentarios
Greg Heath
el 23 de Abr. de 2014
Again: You only have ONE sample of data. You have 3 classes with 1 30-dim example in each class. 3 examples define a 2-dimensional space. Therefore you should use dimensionality reduction and/or try to find more examples. Your design example is really too simple for any thing except a linear, quadratic or nearest neighbor classifier.
For more appropriate design examples see the MATLAB data base
help nndatasets
doc nndatasets
Más respuestas (0)
Ver también
Categorías
Más información sobre Sequence and Numeric Feature Data Workflows en Help Center y File Exchange.
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!