What I'm interested in
I'm mostly writing about various topics in machine learning. The posts here are the result of the notes I write as I dig into these topics ...
Matching Flows and Scores
Flow Matching, Diffusion, Scores and Characteristic Functions
Doob's h-Transform
Wrong Direction, buddy
Girsanov's Theorem
Walk the Walk (differently)
The Adjoint Method for Neural ODE's
Two perspectives on Memory Efficient Gradients
Guten Tag, Herr Ehrenfest. You seem to be living in two worlds?
The Link between Discrete and Continuous Diffusion
Black-Scholes Equation for Options Pricing
Optional Derivatives
Simple Reverse-Time SDE Derivation for Diffusion Models
Save yourself a lot of Bayes with a linear function
Fokker, Planck & Ito
Fokker-Planck Equation Via Ito Calculus
Fokker, Planck & Kolmogorov Revisited
Distributions as partial differential equations over time
Fast Fourier Transform
From Complex Exponentials to Frequencies in O(N log N)
Continuous Time Markov Chains in Jax
jax.jit(jax.vmap(x=Discrete Space, t=Continuous Time))
Stein's Lemma for Trace Estimation
Warning: May contain traces of nuts (and matrices)
RSA Cryptography:
The Math
The math that ensures that it's none of your business
Reverse Time Stochastic Differential Equations [ for generative modelling ]
'If I Could Turn Back Time' by Cher (1989)
Bernoulli to Poisson Point Processes
From coin flips to stochastic processes
(Importance Weighted) Variational Autoencoders Derived
How to get to the objective function of VAEs ...
The Adjoint Method in Neural Ordinary Differential Equations
Reverse-Mode Sensitivity Training
Solving (Some) SDEs
Geometric Brownian Motion & Ornstein-Uhlenbeck Process
Ito's (Di)Lemma
Or how to differentiate a function of a stochastic process.
Stochastic Differential Equations
A Wiener twist to differential equations
Intro to Bayesian Inference
Do you have a minute to talk about our lord and saviour, Thomas Bayes ... ?
(Basic) Inducing Points in Gaussian Processes
Tackling the computational cost of GP's
Sinkhorn Iterations
Computing Wasserstein Distances
GPyTorch
GPU Acceleration for Gaussian Processes
Hamiltonian Mechanics in Monte Carlo Samplers
Sampling with the help of physics
Bayesian Non-Negative Matrix Factorization
Being Unsure About What To Recommend
The Reparameterization Trick
Outsourcing Stochasticity and Making Normal Distributions Differentiable
Non-Negative Matrix Factorization
What Shall We Recommend?
Easter Walk Poem
Goethe's Easter Walk, the Berlin Version
Bayesian Optimization
Using Gaussian Processes for Optimization
Gaussian Processes - Extensions
Extensions to Gaussian Processes
Gaussian Processes - Basics
A Tutorial for Gaussian Processes
RL in Challenging State Spaces and Sparse Reward Signals
Talk on AlphaGo and Hierarchical RL
Introduction to Reinforcement Learning
A Tutorial Talk on RL