Evolutionary Optimization Trumps Adam Optimization on Embedding Space Exploration
This work addresses the challenge of controllability in deep generative models for image generation, offering an incremental improvement through gradient-free optimization.
The paper tackled the problem of optimizing image generation in diffusion models without expensive retraining by comparing evolutionary optimization (sep-CMA-ES) with Adam on Stable Diffusion XL Turbo's prompt embeddings, finding that sep-CMA-ES consistently yields superior improvements in aesthetic and alignment metrics on the Parti Prompts dataset.
Deep generative models, especially diffusion architectures, have transformed image generation; however, they are challenging to control and optimize for specific goals without expensive retraining. Embedding Space Exploration, especially with Evolutionary Algorithms (EAs), has been shown to be a promising method for optimizing image generation, particularly within Diffusion Models. Therefore, in this work, we study the performance of an evolutionary optimization method, namely Separable Covariance Matrix Adaptation Evolution Strategy (sep-CMA-ES), against the widely adopted Adaptive Moment Estimation (Adam), applied to Stable Diffusion XL Turbo's prompt embedding vector. The evaluation of images combines the LAION Aesthetic Predictor V2 with CLIPScore into a weighted fitness function, allowing flexible trade-offs between visual appeal and adherence to prompts. Experiments on a subset of the Parti Prompts (P2) dataset showcase that sep-CMA-ES consistently yields superior improvements in aesthetic and alignment metrics in comparison to Adam. Results indicate that the evolutionary method provides efficient, gradient-free optimization for diffusion models, enhancing controllability without the need for fine-tuning. This study emphasizes the potential of evolutionary methods for embedding space exploration of deep generative models and outlines future research directions.