Please solve the areas where “Pass” is mentioned using Python 3.9.7…
Question Answered step-by-step Please solve the areas where “Pass” is mentioned using Python 3.9.7… Image transcription textSVM and RF error rate vs. fraction of training data used SVM 0.50 RF 0.45 0.40 e: Testing error rate 0.35 0.300.25 0.20 10-2 10-1 100 f: Fraction of available training data used Sample output of average SVM, RF errorrates vs. fraction of training data… Show more Please solve the areas where “Pass” is mentioned using Python 3.9.7import numpy as npimport matplotlib.pyplot as pltfrom sklearn import datasets, svmfrom sklearn.ensemble import RandomForestClassifier”””ASSIGNMENT: COMPLETE err_test function below””””””This code plots SVM and RF error rates vs. size of training set””””””def random_sample(N, n):return a random sample of [0,…,N-1] of size n”””def random_sample(N, n): return np.random.choice(N, n, replace=False)”””def err(cls, X_tr, y_tr, X_te, y_te, cls):train classifier cls and return error rate”””def err(cls, X_tr, y_tr, X_te, y_te): cls.fit(X_tr, y_tr) z_te = cls.predict(X_te) return len([1 for y, z in zip(y_te, z_te) if y != z])/len(y_te)”””COMPLETE:err_test(M, f_all, X_tr, y_tr, X_te, y_te)conducts M Monte Carlo trials for both the SVM and RF classifier for each ff in f_all is the fraction of available training data to use on the classifierX_tr, y_tr is the whole set of available training data: fraction f is usedX_te, y_te is the set of testing data available for measureing performanceThe code should return two dictionaries: e_svm_ave, e_rf_aveBoth dictionaries have keys that are the training data fractions f_allThe values for key f is the average classifier error (over the M trials) for f”””def err_test(M, f_all, X_tr, y_tr, X_te, y_te): # initialize empty dictionaries for SVM and RF average error rates # named e_svm_ave and e_rf_ave # dictionary keys are the fractions in f_all # dictionary values are the average error for that fraction of training data e_svm_ave, e_rf_ave = {}, {} # iterate over each fraction f in f_all for f in f_all: # 1. n_tr is the number of training points to use with fraction f # note len(y_tr) is the total number of training points available pass # 2. initialize empty lists for SVM and RF errors # named e_svm and e_rf # these will hold the M sample error rates for the M Monte Carlo trials pass # conduct M Monte Carlo trials for _ in range(M): # 3. generate random set of indices i_tr of size n_tr # make sure to use the random_sample function defined above # let i_tr be the random indices returned from random_sample pass # 4. use indices i_tr to specify the *actual* training set # let X_tra, y_tra be actual training set, sampled from X_tr, y_tr # using random indices i_tr pass # 5. append to e_svm the value returned by the err function above # where the classifier is the SVM classifer svm.SVC() # and using X_tra, y_tra, X_te, y_te pass # 6. append to e_svm the value returned by the err function above # where the classifier is the SVM classifer RandomForestClassifier() # and using X_tra, y_tra, X_te, y_te pass # 7. update dictionaries with average Monte Carlo trial error rates # use np.mean() on e_svm and e_rf to get Monte Carlo average # this is the value in the e_svm_ave and e_rf_ave dictionaries # for the key f pass # return SVM and RF average error rate dictionaries return e_svm_ave, e_rf_ave”””def plot_err(e_svm, e_rf, fn):plot the SVM and RF error rates vs. fraction of used training data”””def plot_err(e_svm, e_rf, fn): plt.scatter(e_svm.keys(), e_svm.values(), label=’SVM’) plt.scatter(e_rf.keys(), e_rf.values(), label=’RF’) plt.xscale(‘log’) plt.title(‘SVM and RF error rate vs. fraction of training data used’) plt.xlabel(‘f: Fraction of available training data used’) plt.ylabel(‘e: Testing error rate’) plt.legend() plt.savefig(fn)if __name__ == “__main__”: # load the covertype dataset # https://scikit-learn.org/stable/modules/generated/sklearn.datasets.fetch_covtype.html # https://scikit-learn.org/stable/datasets/real_world.html#covtype-dataset dataset = datasets.fetch_covtype() # get feature and target data X, y = dataset.data, dataset.target # original and target sample size N, n = X.shape[0], 10000 # random indices for sampling data to more manageable size i_sd = random_sample(N, n) # use the random indices to downsample X, y X_sd, y_sd = X[i_sd], y[i_sd] # target testing size n_te = 1000 # random indices for testing data i_te = random_sample(n, n_te) # use the random indices to set testing data X_te, y_te X_te, y_te = X_sd[i_te], y_sd[i_te] # i_tr are indices in range(n) not in testing indices i_te i_tr = list(set(range(n)) – set(i_te)) # use the indices to set the (max possible) training data X_tr, y_tr X_tr, y_tr = X_sd[i_tr], y_sd[i_tr] # fraction of available training data to use # recall: np.geomspace gives points evenly distributed on a log-axis f_all = np.geomspace(0.01, 1.00, 20) # number of Monte Carlo independent trials (for averaging out errors) M = 10 # average error rates for SVM and RF for each fraction f e_svm, e_rf = err_test(M, f_all, X_tr, y_tr, X_te, y_te) # print error rates for both SVM and RF for each fraction f print(“fte_svmte_rf”) [print(“{:.3f}t{:.3f}t{:.3f}”.format(f, e_svm[f], e_rf[f])) for f in f_all] # plot error rates and save to file fn fn = “SampleOutput.pdf” plot_err(e_svm, e_rf, fn) Computer Science Engineering & Technology Python Programming ECE 105 Share QuestionEmailCopy link Comments (0)
