Different values for SVM parameters on different runs of grid search on LibLinear

Hi,

I am trying to fine tune my parameters in a LinearSVM using libLinear via scikits-learn package. I observe different values of C on different runs on my program which affects the accuracy of my classifier and results in variations. Here's what I do :

    X = sparse_feats
    Y = target_labels

    train, test = iter(StratifiedKFold(Y, 2, indices = True)).next()
    C_start, C_end, C_step = -3, 15, 2

    # Generate grid search values for C
    C_val = 2. ** np.arange(C_start, C_end + C_step, C_step)
    print C_val
    grid_clf = svm.sparse.LinearSVC()        
    print grid_clf

    linear_SVC_params = {'C': C_val}

    grid_search = GridSearchCV(grid_clf , linear_SVC_params, n_jobs = 10, iid = False, score_func = f1_score)
    grid_search.fit(X[train], Y[train], cv=StratifiedKFold(Y[train], 10)) 
    y_true, y_pred = Y[test], grid_search.predict(X[test])

    print "Classification report for the best estimator: "
    print grid_search.best_estimator

    print "Tuned for  with optimal value: %0.3f" % f1_score(y_true, y_pred)
    print classification_report(y_true, y_pred)

    print "Grid scores:"
    pprint(grid_search.grid_scores_)

    print "Best score: %0.3f" % grid_search.best_score

    best_parameters = grid_search.best_estimator._get_params()

I am not sure why I get different values of C on different runs of my program. I initially felt it may be due to the randomness of the StratifiedKFold function however, I manually sliced the numpy arrays into halves and tried things out but the variations still persist. The C value range is taken from the libSVM guide.