cvbb

Boundary of DT/NN-HOW many layers
                       SVM- visual data(which gives 0 error)
MLE- solve parameters
Naive Bayees

LOG.REG-  Equation solving
        find decision boundary
         Boundary of logistic classifier

MODEL Evaluation- def acc,prec,recall

k- fold cross validation
       (which gives lower error)

ROC Curve(TPR Vs FPR)-Area And its dependency

KNN with k-fold  , what is the decision boundary of 3,4 etc

Ensemble- Bias- Var tradeoff
      effect of increasing number of layers in NN
      effect of adding more variables to a model
      effect of increasing no.of trees in random forest
      effect of increasing K in KNN
     making DT instead of stump as base classifier

bootstrapping, boosting(adaboost)
                 no.of iterations of adaboost needed
                  epselon, gamma new weights given no.of  misclassified points

K- means
       assignment, update, centres
       strength and weakness
       heirarchical clustering
       Single link, max link......
       EM soft clustering

PCA- what does it do

Bayesnet
    no.of parameters needed- each have 4 choices
    compute joint probability- factoring
    Dependency Separation

HMM
    Forward algorithm
    Viterbi- healthy, fever

RNN
     Q,V function- dependency on policies







Questions

no.of parameters- sequential, bayes net

adaboost- error calc, weight update
 1st iter-normalized weights
 2nd- ist and 2nd iter
 adaboost practice questions
    anjum chida slides- heart shape problem

probability-office, light, computer on, login- online office light probability-0.4 near

knn,ann,perceptron,naive bayees-acendining order of speed model creation

No.of parameters-match the following-all independent(x1-xn), bayees network description, all dependent, sequential order

sigmoid, derivative graphs -100 to 100, -1 to 1

viterbi- health fever
 diagnostic queries- given cough----temparature HHF

forward- 2 states(o1,o2 observebles probability)

5 options- which cases reduces variance-knn-ann-decision tree

Given a constant function-variance high or low

	another question similar to variance and bias

	variance- underfit/overfit
    	decision tree - depth dependency
    	Ann- no.of layers
    	Z=x+y variance of Z x (0-1),y(0-2) uniform 5/2,1/2,/4,/3

logistic regression-- x<6 y==1/0 quiz question same

MLE-P(X) theta

MLE- sigma xi/5n in denominator

dataset-which classifier gives zero training error(2 q)

k-means clusters- sum of distances(cluster point to centroid)
                  centroids after each iteration
                  Fails in which case.
	k means-  k-1, k-3 error
	k- means fail cases
           outlier,spherical, centroid calculations


Bayes net-joint probability(p-----)
          deseparation(not conditionally independent select)
	  d-separation bayes net- conditional independence from the diagram earth quake,call - middle and last

graph points 7-svm with 7 fold----> training error