def output_loo(X, y, n_features=range(5, 31, 5)):
    scaling = preprocessing.StandardScaler()
    scaling2 = MinMaxScaler()

    svml1 = svm.LinearSVC(C=0.5, loss='l2', penalty='l1', dual=False)
    lrl1 = LogisticRegression(C=1, penalty='l1', tol=0.4)
    svml2 = svm.LinearSVC(C=0.5, loss='l2', penalty='l2', dual=False)
    lrl2 = LogisticRegression(C=1, penalty='l2', tol=0.4)
    svml = svm.LinearSVC()
    lrl=LogisticRegression()

    cv=LeaveOneOut(len(y))
    all_loo=[]
    nf=[]

    for num in n_features:
        allchi2 = SelectKBest(chi2, k=num)
        my_svm11 = Pipeline([('scaling', scaling2), ('chi2', allchi2), ('svml1', svml1)])
        my_svm21 = Pipeline([('scaling', scaling2), ('chi2', allchi2), ('svml2', svml2)])
        my_svm = Pipeline([('scaling', scaling2), ('chi2', allchi2), ('svm', svml)])

        my_lr11 = Pipeline([('scaling', scaling2), ('chi2', allchi2), ('lrl1', lrl1)])
        my_lr21 = Pipeline([('scaling', scaling2), ('chi2', allchi2), ('lrl2', lrl2)])
        my_lr = Pipeline([('scaling', scaling2), ('chi2', allchi2), ('lr', lrl)])

        loo=[]
        loo.append(skl.cross_validation.cross_val_score(my_svm11, X, y, cv=cv, scoring="accuracy").mean())
        loo.append(skl.cross_validation.cross_val_score(my_svm21, X, y, cv=cv, scoring="accuracy").mean())
        loo.append(skl.cross_validation.cross_val_score(my_svm, X, y, cv=cv, scoring="accuracy").mean())
        loo.append(skl.cross_validation.cross_val_score(my_lr11, X, y, cv=cv, scoring="accuracy").mean())
        loo.append(skl.cross_validation.cross_val_score(my_lr21, X, y, cv=cv, scoring="accuracy").mean())
        loo.append(skl.cross_validation.cross_val_score(my_lr, X, y, cv=cv, scoring="accuracy").mean())

        all_loo.append(loo)
        nf.append(str(num))
    res=np.round(np.array(all_loo), 3)
    colnames=['SVM with L1', 'SVM with L2', '  SVM  ', 'LR with L1', 'LR with L2', '  LR  ']
    restable=pandas.DataFrame(res, index=nf, columns=colnames)

    return restable