Fast and Lightweight Multiagent Image Generation Based on the Evolution Methods

The paper presents a novel approach to the multiagent image generation leveraging the evolutionary methods to enhance both speed and efficiency. The developed method utilizes latent noise as an input to generate the facial images achieving a 10-fold increase in the learning speed while maintaining the high-quality output. Our approach integrates the lightweight multiagent methods with facilitating rapid adaptation in the complex environments. Experimental results demonstrate significant improvements in the learning time and image quality and demonstrate the evolution-driven techniques evolution in the efficient distributed generative tasks.