Featured
Defusing Diffusion Models
How not to get a headache working with diffusion models.
Diffusion models are all the rage these days. They are a new class of generative models, which are capable of generating high-quality images. They have attracted a lot of attention...
All Stories
Hodgkin-Huxley Model
Quantifying the activity of a neuron
Back in 1952 Hodgkin and Huxley found that there are three main types of currents describing the dynamics of a neuron: Sodium (N), Potassium (K), and a leak current, which...
Langevin Monte Carlo: Sampling using Langevin Dynamics
How to sample from a distribution using Langevin Dynamics and why it works.
Generative models rely on the ability to sample from a distribution. When the distribution is known the task of sampling from it is straightforward, for example, sampling from a normal...
What is linear regression?
On fitting a line to your data
Linear Regression (LR) is one of the most fundamental algorithms in machine learning. It is the simplest regression model out there it is very often a fitting solution for a...
AI learns to sail upwind
Solving sailing using Q-Learning
The year 2020 was not the greatest for sailors. The pandemic limited the available options for voyages or even to gather a crew. I am a huge sailing enthusiast and...
What is a Loss Function?
How to formulate the right objective.
Any machine learning student will learn about loss functions sooner rather than later. They are a fundamental element of learning and optimisation, therefore understanding is necessary for mastering machine learning....
