DELHI DERVILES:

CHENNAI MUMBAI DELHI BANGLORE RAJASTHAN PUNJAB HYDRABAD KOLKATTA

DELHI:

Coach: South Africa Ray Jennings Bowling coach:India Venkatesh Prasad
Captain:New Zealand Daniel Vettori Colours: RCBFounded: 2008
Home ground:M. Chinnaswamy Stadium(Capacity: 40,000) Owner:Dr.Vijay Mallya
Chief executive:Brijesh Patel Manager:Avinash Vaidya
Adviser:Anil Kumble First-class debut:Kolkata Knight Ridersin 2008at M. Chinnaswamy Stadium
Official website:Royal Challengers Bangalore

PCA and its Variants:
PCA implements the Karhunen-Loève (KL)transform, a classical unsupervised second-order
method that uses the covariance matrix’s eigenvaluesand eigenvectors to transform the feature
space, creating orthogonal uncorrelated features.We use five different variations of the KL
transform: KL1 (the original PCA) through KL5.The difference between each is how we choose thecomponents that will compose the feature vectorsand how we incorporate additional informationfor each class. For instance, the eigenvectors aren’tcalculated from the mixture’s covariance
matrix,of each class (Sw). Alternatively, we can order thefeatures according to an entropy-based criterion.The main advantage of these PCA variants overindependent componentanalysis (ICA) is thatthey let us incorporate statistical knowledge fromeach class (such as class conditional means, variances,and covariances) in the feature-extractionstep and thus define supervised methods. (Moredetails about PCA and its variants are available.

Classifier Combination
Multiclassifier fusion integrates multiple classifierstogether to improve classification accuracy.
Researchers have successfully adopted thismethod in many pattern recognition applications,including audio classification, handwriting recognition,and speech recognition. However, classifiercombinations for face recognition problemshave yet to be deeply investigated. The reasonthis approach improves classification performanceis simple: each individual classifier is supposed toproduce classification errors that are assumed tobe uncorrelated.7 Therefore, combining classifiersthat are complementary to each other could helpthe entire system capture additional informationunachievable by a single classifier. As a result, thisreduces the overall Basically, when building an ensemble of classifiers,we can vary four levels:
• the combination level, using different combination
strategies;
• the classifier level, usin

CHENNAI MUMBAI DELHI BANGLORE RAJASTHAN PUNJAB HYDRABAD KOLKATTA