Flow Matching

KL-Divergence flow-matching Generative models are very good / best in class approximators of a complex probabilistic equation/distribution that we most of the times have no idea of !! Images modality : All images in the world are from a very complex distribution of pixels that gives direction , based on the prompts , and are dependent of what the user wants , and as it very complex to model that distribution we rely on NN to predict it , hence we have diffusion models ...

The pre-requisites for this blog post is Journey till diffusion model Stable Diffusion model: Uses cross-attention for allowing conditional modelling ( using text/segmentation map + image to generate image) Mode collapse doesnt happen in likelihood based model and SD is a likelihood based model High frequency details : it means the details / detail-oriented view Related work: (previous work in this field) , same vq-vae and vq-gna What is the Inductive bias of the DM’s inherited by the UNET model ? ...