Step $1$: use assignment $4$ draft.py to draw the following six scatter plots depicting the way different classifiers

perform on the iris dataset which has $50$ samples for each of its three classes$($labels$).$ These plots figure each

class with a different color $($red$,$ blue, white$)$

PCA.f$0$

PCA.f$0$

PCA.f$0$

SVC w$/$ Sig. $6$

score: $65\mathrm{.}2\%$

PCA.f$0$

PCA.f$0$

Step $2$: $(25$ points$)$ Compare six SVM classifiers with polynomial kernel $($kernel$=$'poly'$),$ degrees $[3,5,7],$ and

gamma values $[\mathrm{.}3,\mathrm{.}5]$ by drawing a grid of six plots similar to the one you obtained from step $1.$

Step $3$: $(25$ points$)$ Compare the following six classifiers by drawing a grid of six plots similar to the one you

obtained from step $1$:

linear regression

linearSVC

GaussianNB w$/$ var$\_$smoothing$=2$e$1$

GaussianNB w$/$ var$\_$smoothing$=1$e$1$

GaussianNB w$/$ var$\_$smoothing$=1$e$0$

GaussianNB w$/$ var$\_$smoothing$=1$e$-1$

Step $4$: $(25$ points$)$ Compare six variations of SVM classifiers with kernels $[\text{'}$sigmoid$\text{'},\text{'}$rfb$\text{'}],$ and gamma values $[1$e$-$

$1,1$e$0,1$e$2]$ by drawing a grid of six plots similar to the one you obtained from step $1.$

Step $5$: $(25$ points$)$ Compare eight variations of neural network MLP classifiers with default alpha $(1$e$-4),$ solvers

$[\text{'}$adam$\text{'},$ 'Ibfgs'$],$ activation $[\text{'}$logistic$\text{'},$ 'relu'$]$ and layers $[(30,30),(10,5)]$ by drawing a grid of eight plots similar to the

one you obtained from step $1.$

assignment $4$ draft.py

from sklearn.decomposition import PCA import matplotlib.pyplot as plt from sklearn import svm$,$ datasets from sklearn.naive$\_$bayes import GaussianNB from sklearn.inspection import DecisionBoundaryDisplay from sklearn.model$\_$selection import train$\_$test$\_$split from sklearn.neural$\_$network import MLPClassifier from sklearn.linear$\_$model import LinearRegression pca $=$ PCA$(2)$ #Step $1$: loading the dataset iris $=$ datasets.load$\_$iris$()$ #Step $2$: transforming the dataset features to reduce dimensions from $4$D to $2$D X$\_$transformed $=$ pca.fit$\_$transform$($iris$.$data$)$ #Step $3$: obtaining the true labels of dataset y $=$ iris.target #Step $4$: splitting the dataset into $15\%$ testing set and $85\%$ training set in a random fashion X$\_$train, X$\_$test, y$\_$train, y$\_$test $=$ train$\_$test$\_$split$($X$\_$transformed, y$,$ test$\_$size$=0\mathrm{.}15,$ random$\_$state$=23)$ #Step $4$: Defining $5$ models and fitting them to our training set models $=($ LinearRegression$(),$ svm$.$LinearSVC$($C$=1,$ max$\_$iter$=10000),$ svm$.$SVC$($kernel$=$'poly', degree$=3,$ gamma$=\mathrm{.}1,$ C$=1),$ MLPClassifier$($solver$=\text{'}$lbfgs$\text{'},$ alpha$=\mathrm{.}5,$ hidden$\_$layer$\_$sizes$=(2,2),$activation$=$'logistic', random$\_$state$=23),$ GaussianNB$($var$\_$smoothing$=1$e$1),$ svm$.$SVC$($kernel$=$"sigmoid", gamma$=0\mathrm{.}6,$ C$=1),)$ models $=($clf$.$fit$($X$\_$train, y$\_$train$)$ for clf in models$)$ # Step $5$: Drawing the plots titles $=("$Linear regression", "LinearSVC", $"$SVC w$/$ poly$3\mathrm{.}1","$NN layer:$(2,2)",$ "GaussianNB", $"$SVC w$/$ Sig$\mathrm{.}6")$ fig, sub $=$ plt$.$subplots$(2,3)$ plt$.$subplots$\_$adjust$($wspace$=0\mathrm{.}4,$ hspace$=0\mathrm{.}7)$ X$0,$ X$1=$ X$\_$train$[$:$,0],$ X$\_$train$[$:$,1]$ for clf$,$ title, ax in zip$($models$,$ titles, sub.flatten$())$: disp $=$ DecisionBoundaryDisplay.from$\_$estimator$($ clf$,$ X$\_$train, response$\_$method$=$"predict", cmap$=$plt$.$cm$.$coolwarm, alpha$=0\mathrm{.}8,$ ax$=$ax$,$ xlabel$=$'PCA.f$0\text{'},$ ylabel$=$'PCA.f$1\text{'})$ ax$.$scatter$($X$0,$ X$1,$ c$=$y$\_$train, cmap$=$plt$.$cm$.$coolwarm, s$=20,$ edgecolors$="$k$")$ ax$.$set$\_$xticks$(())$ ax$.$set$\_$yticks$(())$ ax$.$set$\_$title$($title$+\text{'}$

score: $\text{'}+$str$(100*$round$($clf$.$score$($X$\_$test,y$\_$test$),3))+\text{'}\%\text{'})$# plt$.$show$()$