<p>This paper presents a detailed study of a new regularization method based on optimization for the inverse sample covariance matrix. This method is highly computationally efficient and does not require complex computing resources. It is designed for linear receivers in multi-user communication systems with a large number of antennas and operates under conditions of limited sample data. The analysis showed that the probabilistic noise distributions do not significantly affect the optimal value of the regularization factor, which confirms the universality and reliability of the proposed approach. Simulation results demonstrate the superiority of this method over traditional approaches. In particular, the method provides better conditionality of the sample covariance matrix and significantly reduces the computational complexity of calculating the linear equalizer weight matrix in the uplink of a communication system, which makes it especially useful for modern communication systems with high user density and limited computing resources.</p>