\emph{KPCA}
KPCA
-
Related work: Mainstream problems and methods
Preliminaries: Data model, Problem statement
Basis: Theory, Methods (GAN) -
The proposed algorithm
Network architecture (overview)
Algorithm framework
Pseudo code -
why, what, how, when, where
Why augment: small data
How augment: GAN
How to handle small data through GAN:
How to handle imblance data while augment:
What should be kept while augment: positive/negative ratio, ratio equal to 1 (from imbalance to balance, consider cost)
How to ensure the quality of the augmented data: add noise, transpose, overlap, crossover. What type of noise?
What is the quality measure of the data: diversity, clean, ?
Motivation? Scheme? Concrete objectives?
better classification models on this basis.
→
\to
→
good classification performance.
KPCA is used to reduce the feature dimension to obtain the most effective feature space for classification.
→
\to
→
KPCA is used to construct an effective feature space.
We design a KPCA based approach to construct an effective feature space for the microbial data.
What is the relationships between features of microbial data?
Which parameters?
How to set?
Reason!