RSSAmplifier

Blog

christopher beckham

Academic blog: generative models, diffusion, VAEs, and model-based optimisation, amongst other topics. ML researcher

beckham.nzRSS feed ↗10 posts

Latest posts

reflecting on past work – offline model-based optimisation

reflecting on past work -- offline model-based optimisation

paper thoughts – questionable research practices (QRPs) in machine learning

paper thoughts: questionable research practices (QRPs) in machine learning

EDM diffusion models - a Jupyter implementation, and how they are implemented in practice

EDM diffusion models - a Jupyter implementation, and how they are implemented in practice

A deep dive into conditional variational autoencoders

A deep dive into conditional variational autoencoders

Techniques for label conditioning in Gaussian denoising diffusion models

Techniques for label conditioning in Gaussian denoising diffusion models

Learning the conditional prior over classes for image diffusion

\(\newcommand{\xx}{\boldsymbol{x}}\) \(\newcommand{\yy}{\boldsymbol{y}}\) \(\newcommand{\pt}{p_{\theta}}\) \(\newcommand{\QQ}{\boldsymbol{Q}}\) \(\newcommand{\mm}{\boldsymbol{m}}\) \(\newcommand{\alphabar}{\bar{\alpha}}\) \(\newcommand{\mt}{\mu_{\theta}}\) \(\newcommand{\epst}{\epsilon_{\theta}}\) \(\newcommand{\betatilde}{\tilde{\beta}}\) \(\newcommand{\deltatilde}{\tilde{\delta}}\)…

The obsession with SOTA needs to stop

The obsession with SOTA needs to stop

Towards a more sane Mac OS user experience, and I am late to the party

Towards a more sane Mac OS user experience, and I am late to the party Despite making the transition to Mac OS from Windows as early as 2014, I don’t feel like I have made a particularly strong effort to become a ‘power user’ in the past eight years. By ‘power user’, I mean the kind of person that tweaks every little setting and configuration file and hotkey so that they have an optimised workflow…

My notes on discrete denoising diffusion models (D3PMs)

My notes on discrete denoising diffusion models (D3PMs)

Training GANs the right way

Training GANs the right way