Non-parametric kernel density estimation has a good nature and it contains two forms: fixed and variable window width.Their asymptotic convergence rates of the variance are the same,but biases differentiate a lot.Fixed window width and the balloon estimator have the same bias h2,while Ambramson estimator's is h4.In this paper,we get a good nature of the fixed window width estimator,and calculate the bias of Cauchy density's Ambramson es...