Modelling on Local Optima

Initial Designs for Bayesian Optimisation

Wed, 16 Aug 2023 00:00:00 +0000

index

Bayesian optimisation is a powerful optimisation technique for black-box functions and processes with expensive evaluations. It is popular for hyperparameter tuning in machine learning, but has many real-world applications as well.

Test Functions for Bayesian Optimisation

Sun, 16 Jul 2023 00:00:00 +0000

index

Alternative Surrogate Models for Bayesian Optimisation

Thu, 15 Jun 2023 00:00:00 +0000

index

Acquisition Functions for Bayesian Optimisation

Mon, 01 May 2023 00:00:00 +0000

index

Bayesian optimisation is a powerful optimisation technique for black-box functions and processes with expensive evaluations. It is popular for hyperparameter tuning and model selection in machine learning, but has many real-world applications as well. One of the key components of Bayesian optimisation is the acquisition function, which guides the search process by balancing exploration and exploitation of the search space. In this post, we will dive into the role of acquisition functions in Bayesian optimisation and discuss some popular examples.

Kernels for Gaussian Processes

Sun, 16 Apr 2023 00:00:00 +0000

index

This post takes an extensive look at kernels and discusses the rationales, utility, and limitations of some popular kernels, focusing primarily on their application in Gaussian processes and Bayesian optimisation. Along with the discussion are implementations of the kernels in base R.

Bayesian Optimisation from Scratch in R

Sat, 01 Apr 2023 00:00:00 +0000

index

Bayesian optimisation is a powerful technique for optimising expensive functions or processes. In many applications, such as drug discovery, manufacturing, machine learning, or scientific experimentation, the function or process to be optimised may be time consuming or costly to evaluate. Bayesian optimisation provides a framework for sequential experimentation and for finding optima with as few evaluations as possible.

Bespoke Bayesian Model for Batch Effects in High Throughput Biochemical Assays

Sun, 19 Mar 2023 00:00:00 +0000

Code for this project: github.com/AnHosu/bespoke-bayesian-biochem

batch-effects

In previous studies, we built bespoke Bayesian models to fit observations from a biochemical assay with kinetics that could be represented by the Hill equation. Then we scaled that up to a screening experiment. In those studies, our main goal was to achieve good fits for kinetic parameters.

Bespoke Bayesian Model for High Throughput Biochemical Assays

Sat, 23 Apr 2022 00:00:00 +0000

Code for this project: github.com/AnHosu/bespoke-bayesian-biochem

Bespoke Bayesian Model for Biochemical Assays

Sat, 20 Nov 2021 00:00:00 +0000

Code for this project: github.com/AnHosu/bespoke-bayesian-biochem

I am on a quest to improve the model fitting I do on biochemical assays. For some time, I have had this feeling that I should be able to extract more information from the data gathered in biochemical assays, in particular assays with a high throughput. I have been using classical machine learning techniques and generic fitting and optimisation functions to interpret data from such assays. While this approach works, it also neglects much of the available domain expertise. Many of the underlying biochemical mechanisms are known and I would like my models to take that into account so I get results that are more directly interpretable in the context of the hypothesis that required the assay in the first place. In other words, I want a bespoke model.