Conventional adaptive equalization of data transmission channels requires an initial training period during which a known data sequence is transmitted and properly synchronized at the receiver Transmission of the training sequence is not convenient for multipoint data communication networks. Recently, research is directed towards equalization schemes that operate without initial training sequence. Such schemes are called self-adaptive equalizers(SAE). Most of SAE schemes available in literature need a long time to converge to the optimum equalizer. In this paper we propose an equalization scheme that shortens this time significantly. The proposed scheme is a cascade of two adaptive equalizers. The first one is short and thus it converges rapidly. The partial equalization done by this equalizer is coarse and will not enable reliable recovery of data. The second equalizer is long and it starts adaptation after the convergence of the short one. Due to the partial equalization performed by this latter the convergence time decreases. The length of the long equalizer is determined by the desired precision of equalization. This scheme is simulated on a digital computer. Simulation results exhibit superiority of the proposed scheme over published ones.