Printing applications require to convert RGB displayable pictures into four printing process components: yellow, magenta, cyan and black. The printing process generates black color by two mechanisms: substractive mixture of YMC components and by K component. In almost all cases, the RGB picture on the display has different colors to the YMCK converted printed picture. The colors of the YMCK printed picture can be simulated on the RGB display. Simulation is a complex process, which depends on the printer (ink, paper, printing technique, dot gain, UCR or GCR corrections), monitor (RGB phosphor components, gamma correction, brightness and saturation adjustments) as well as observation conditions (illuminant, reflections).In the paper, a neural network is proposed as an alternative solution for RGB-YMCK color conversion, in order to obtain closer color appearance between RGB image and the corresponding YMCK printed image. The YMCK data, as inputs, and the RGB data resulted from simulation of YMCK printed colors, as outputs, are used to learn the neural net-work in order to perform the global color conversion from RGB to YMCK. The general RGB simulation process of the printed YMCK colors is not bidirectional, so that, the network finds one possible transformation with a certain probability, strongly dependent on the learning data which determines the weights of the neural network.