We use cookies in order to improve the quality and usability of the HSE website. More information about the use of cookies is available here, and the regulations on processing personal data can be found here. By continuing to use the site, you hereby confirm that you have been informed of the use of cookies by the HSE website and agree with our rules for processing personal data. You may disable cookies in your browser settings.

  • A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

HDI Lab seminar 'Amortising Intractable Inference with Diffusion Models and Off-policy RL'

12+
*recommended age
Event ended

On August 15, 2024, Nikolay Malkin (University of Edinburgh) will speak on 'Amortising Intractable Inference with Diffusion Models and Off-policy RL'.

Abstract:

I will present recent and ongoing work relating diffusion models, variational inference, and reinforcement learning. Efficient inference of high-dimensional variables under a diffusion model prior enables solutions to problems such as conditional generation, semantic segmentation, and combinatorial optimisation. A new family of amortised methods places the problem of stochastic continuous control -- including sampling of posterior distributions under neural SDE priors -- in the theoretical framework of off-policy entropy-regularised reinforcement learning, via the formalism of continuous generative flow networks. This connection allows us to train diffusion models to sample a target density or energy function, i.e., to perform black-box variational inference. It further lets us fine-tune pretrained diffusion models in a data-free manner for asymptotically unbiased posterior sampling, with applications to vision (class-conditional generation, text-to-image generation), language (constrained generation in diffusion language models), and control/planning (KL-constrained policy extraction with a diffusion behaviour policy). These approaches are applicable in Bayesian machine learning and various scientific domains.

Diffusion models as plug-and-play priors

Improved off-policy training of diffusion samplers

Amortizing intractable inference in diffusion models for vision, language, and control

A theory of continuous generative flow networks

Time: August 15, 14:40

The event will be held online. The get the Zoom link, please contact Elena Alyamovskaya