The paper introduces a general structure for parameter adaptation/learning algorithms (PALA). Thisstructure is characterized by the presence of an embedded ARMA (poles-zeros) filter in the PALA. Thekey question is how to select the coefficients of thisfilter in order, on the one hand, to guarantee the stability of the parameter estimator for any (positive)value of the adaptation gain/learning rate and for anyinitial conditions and on the other hand to acceleratethe adaptation transient. In order to achieve this, itis shown that on one hand the embedded ARMA filter should be characterized by a positive real transferfunction and on the other hand the filter acting onthe correcting term (the dynamic adaptation gain)should be characterized by a strictly positive realtransfer function. Specic conditions for the designof a second order ARMA embedded filter (ARIMA2algorithm) are provided.It is shown in the paper that many parameter adaptation/learning algorithms (PALA) used inadaptive control, system identication and neural networks (Nesterov, Conjugate gradients, Momentum back propagation, Averaged gradient, Integral+proportional+derivative, ...) are particularcases of the PALA structure introduced in this paper and specic conditions for the stable operation ofthese algorithms are given.Performance of the ARIMA2 algorithm as well as ofthe other algorithms reviewed in the paper will becomparatively evaluated by simulations and experimental results obtained on an active noise controlsystem.