Introduction

This post is the first of two about how I've trained a Latent Diffusion Model (LDM) to generate images from ImageNet-1k class labels.

As a recap from the last post, which was a high level overview of how a LDM works we have two main components:

1. A Variational Autoencoder …

Motivation

A colleague recently asked me what problems I enjoy working on, so I've been thinking more and more about where I draw enjoyment from, especially in my diffusion model project, but also from other previous projects and what I have worked on for others. This post is to organise …

Project

This project builds a Latent Diffusion Model (LDM) on CIFAR-10 that has been upscaled to 64x64.

Introduction

After exploring diffusion on MNIST datasets, I have continued exploring more, successfully building a LDM on CIFAR-10 data. This post covers the first phase of a currently three month project helping to …

The Problem: TSP

The Traveling Salesperson Problem (TSP) is one of the most famous combinatorial optimisation problems. It is incredibly easy to state: Given a list of cities and the distances between each pair, find the shortest route that visits each city exactly once and then returns to the first …

Demo

Before I even get started on what I've done, here is a demo that you can play with.

Introduction

After playing around with reconstruction and classification of MNIST digits last week and talking to a few colleagues about the project, I decided to investigate diffusion models. I broadly knew …

Introduction

This blog post is about my attempts to learn about self-supervised learning. Most of my background is in reinforcement learning as opposed to any kind of supervised learning, so I feel I need to broaden my knowledge and be aware of how to actually structure projects, issues that can …