LT – Lara Thompson:: MLer etc

Studied physics, math, ~neuroscience; now do ML.

Boulder and run, read a lot, write some, always learning.

ML because it felt natural after physics, closer to what matters in the day to day, exciting in how powerful it can be, challenging in how badly it can fail; mixing math, programming, and problem solving to create products and services that help people rather than manipulate or distract them, and build technology that means less waste instead of yet more stuff.

2024 Update

best laid plans…

Random list of updates and imminent plans.

Improving Training Stability for Multitask Ranking Models in Recommender Systems

eg. Making it easier to train complex models fast

Notes for a Data Science Meetup, background info for the Clippy paper.

Embeddings

They aren’t just for text

Some introductory notes on embeddings as pre-reading for our data science paper reading group.

How Bad is Top-K Recommendation under Competing Content Creators?

Introduction for Newbies and Summary for the Rushed

Backround notes on recommenders from a game theory / bandit learning perspective to help people understand this week’s paper.

Notes on Simple Embodied Language Learning as a Byproduct of Meta-RL

Notes for the Data Science Meetup discussing meta-RL model DREAM.

VALL-E Notes

Notes for the Data Science Meetup discussing VALL-E.

Assess Yourself for Better Bouldering

Data-Backed Training Advice

Analysis of a climbing survey dataset to test the wisdom of the crowd in how best to improve at climbing and assess training progress.

Notes on ETA Prediction with GNNs in GMaps

Background and paper notes for the Data Science Meetup.

Data is everywhere!

Uh, where again?

Last night I was on a data science career panel (of awesome ladies!) as part the Vancouver Datajam 2020 and I promised (as I’ve been meaning to do for a while…) to post a list of data resources.

Linear Regression with Uncertainty

Everyone uses linear regression whether they admit it or not. Here I review the math behind it with estimates of uncertainty.

References for Data, ML, Math etc.

Some dated and more or less relevant references to get started in ML.

VanillaNet Notes

Notes for the Data Science Meetup discussing VanillaNet.

Learning to learn without forgetting: a summary

Bullet-point form summary of this excellent paper. Contributions include a memory store of past examples; balances learning of new updates with recall of these stored examples; continual (online) learning in supervised and reinforcement learning settings.

There ain’t no such thing as a free lunch

Notes on the No Free Lunch Theorems for Optimization paper.