sampling – Andrew Fairless, Ph.D.

What I Read: Flow Matching

By Andrew Fairless on June 26, 2024April 23, 2024

https://mlg.eng.cam.ac.uk/blog/2024/01/20/flow-matching.html An introduction to Flow MatchingTor Fjelde, Emile Mathieu, Vincent Dutordoir “Flow matching (FM) is a recent generative modelling paradigm which has rapidly been gaining popularity in the deep probabilisticContinue readingWhat I Read: Flow Matching

What I Read: diffusion distillation

By Andrew Fairless on May 13, 2024March 11, 2024

https://sander.ai/2024/02/28/paradox.html The paradox of diffusion distillationSander DielemanFebruary 28, 2024 “…let’s take a closer look at the various ways in which the number of sampling steps required to get good resultsContinue readingWhat I Read: diffusion distillation

What I Read: Diffusion models, new theoretical perspective

By Andrew Fairless on April 8, 2024March 21, 2024

https://www.chenyang.co/diffusion.html Diffusion models from scratch, from a new theoretical perspectiveChenyang Yuan2023 “This tutorial aims to introduce diffusion models from an optimization perspective…”

What I Read: Sampling Text Generation

By Andrew Fairless on March 19, 2024February 5, 2024

https://huyenchip.com//2024/01/16/sampling.html Sampling for Text GenerationChip Huyen1/15/24 “ML models are probabilistic…. This probabilistic nature makes AI great for creative tasks…. However, this probabilistic nature also causes inconsistency and hallucinations. It’s fatalContinue readingWhat I Read: Sampling Text Generation

What I Read: LLM Training, RLHF

By Andrew Fairless on October 30, 2023October 5, 2023

https://magazine.sebastianraschka.com/p/llm-training-rlhf-and-its-alternatives LLM Training: RLHF and Its AlternativesSebastian Raschka, PhDSep 10, 2023 “RLHF is an integral part of the modern LLM training pipeline due to its ability to incorporate human preferencesContinue readingWhat I Read: LLM Training, RLHF

What I Read: Markov Chains

By Andrew Fairless on October 24, 2023September 9, 2023

https://elevanth.org/blog/2017/11/28/build-a-better-markov-chain/ Markov Chains: Why Walk When You Can Flow?Richard McElreath28 November 2017 “If you are still using a Gibbs sampler, you are working too hard for too little result. Newer,Continue readingWhat I Read: Markov Chains

What I Read: Generative Modeling by Estimating Gradients

By Andrew Fairless on March 7, 2022January 20, 2022

https://yang-song.github.io/blog/2021/score/ Generative Modeling by Estimating Gradients of the Data DistributionAuthor Yang SongDate May 5, 2021 “This blog post gives a detailed introduction to score-based generative models. We demonstrate that thisContinue readingWhat I Read: Generative Modeling by Estimating Gradients

What I Read: MCMC for big datasets

By Andrew Fairless on February 22, 2022January 20, 2022

https://www.pymc-labs.io/blog-posts/pymc-stan-benchmark/ MCMC for big datasets: faster sampling with JAX and the GPUScaling PyMC using JAX to sample on the GPUAuthored by Martin Ingram on 2021-12-22 “You’ll often hear people sayContinue readingWhat I Read: MCMC for big datasets

What I Read: Improving a Machine Learning System (Part 1 – Broken Abstractions)

By Andrew Fairless on December 22, 2021November 16, 2021

https://danshiebler.com/2021-11-06-ml-systems-1/ Improving a Machine Learning System (Part 1 – Broken Abstractions)Dan ShieblerPosted on November 6, 2021 “Making measurable improvements to a mature machine learning system is extremely difficult. In thisContinue readingWhat I Read: Improving a Machine Learning System (Part 1 – Broken Abstractions)

What I Read: AutoML, Multi-task learning, Multi-tower models, Ads

By Andrew Fairless on April 20, 2021April 3, 2021

https://medium.com/pinterest-engineering/how-we-use-automl-multi-task-learning-and-multi-tower-models-for-pinterest-ads-db966c3dc99e How we use AutoML, Multi-task learning and Multi-tower models for Pinterest AdsPinterest EngineeringAug 19, 2020Ernest Wang | Software Engineer, Ads Ranking “In this blog post, we explain how keyContinue readingWhat I Read: AutoML, Multi-task learning, Multi-tower models, Ads

Tag: sampling