Traditional bipolar differential amplifiers have only a ±5mV operational range with nonlinear distortion below 0.1dB. In this paper, a linearization technique based on neural network training algorithm is proposed to expand this 0.1dB linear region to a much wider ±200mV range. Compared with the traditional and recent state-of-the-art techniques for linearization, where gain or noise performance is always being tradeoff for high linearity, the proposed technique leads to the increase of the amplifier's gain with low noise. Another advantage of the proposed approach is that, it is implemented using a simple, paralleled structure that leads to a smaller area and less power consumption than other linearization schemes. The design procedure also enables a completely customizable solution based on required linear range, ripples, and the number of available transistors. The experimental results show that the proposed linearized amplifier has a very good linearity and is capable to wide bandwidth, good gain, and low noise performance.