? I made a pun about the wind but it blows.

Hey nerds,

the weather over here has been incredible this week, so I have been going out as much as possible. Additionally, I have started writing more again, which feels great. But let’s dive into our weekly machine learning:

The Latest Fashion

• Setting up Cuda on new architecture often is a nightmare, [...]
• Inspect intermediate layers of your Pytorch models [...]
• Ever work with tabular data, this paper has done the work in comparing all machine learning models for your convenience.

This is "Light to the Party" all links and extra content can be found in the?full issue from last week. Want the latest in your inbox??Join 333+ other curious minds.

Hot off the Press

Earlier this week, I wrote up my research into navigating a pandemic world in which the responsibility to wear a mask and vaccinate has been passed off to the individual over mask mandates. Here’s the information I found about wearing masks safely and the behavioural measures I distilled.

Two weeks ago, I asked about methods to learn new topics, so I have started writing up some resources to recommend to people trying to learn new topics:

Here are 3 courses to level up your MLOps skills.

Then there are 3 books to learn data analytics.

Finally, some robotics to get into with 3 courses on reinforcement learning.

Machine Learning Insights

Last week I asked, “When would you use the R2 score and how is it defined?” and here’s the gist of it:

The R2-score is also known as the coefficient of determination.

It’s a measure used for regression problems, evaluating how well a function fits a set of data points. Its values range from positive 1 for the best possible score to arbitrary negative values.

The R2-score calculates how well the model explains the variance of the target values based on the input samples. So a model with an R2 of 1 would explain the variance perfectly. Negative values arise when non-linear models represent the output worse than taking the mean of the data.

This score is sometimes preferred over raw mean squared error or MAE, as the R2 has an intuitive interpretation, whereas MSE can achieve arbitrary values that are based on many factors including the range of the output data.

As a bonus, scikit-learn has an easy implementation in sklearn.metrics.r2_score.

Question of the Week

• What are good ways to handle missing data?

Post them on in the comments. I'd love to see what you come up with!

Tidbits from the Web

• If you like StatQuest, you’ll be delighted to hear there’s now [...]
• A tale of dealing with Youtube as a creator: [...]
• Generating a whole 3D landscape in a tweet

This is "Light to the Party" all links and extra content can be found in the?full issue from last week. Want the latest in your inbox??Join 333+ other curious minds.

Vacancies

Should this be a section? Let me know if you’d like job advertisements as they’re submitted to me.

AI and Machine Learning Coordinator

The European Centre for Medium-Range Weather Forecasts (ECMWF) is looking for an AI and ML coordinator. The AI and Machine Learning Coordinator is responsible for leading and coordinating machine learning activities – both scientifically and technically – across all departments and functions at ECMWF, including DestinE. The coordinator will also ensure that the DestinE activities in this area are coordinated between ECMWF and its contractors and between ECMWF and the other entrusted entities in DestinE, namely ESA and EUMETSAT. The candidate will ensure that the roadmap is implemented through the ECMWF 1-year plan and 4-year programme and will regularly review and update the roadmap within the context of the wider ECMWF strategy and the last science-technology developments in the machine learning domain.

Full Description: https://jobs.ecmwf.int/displayjob.aspx?jobid=59

Rate: GBP78,702.36 or EUR 92,770.32?net annual basic salary?+ other benefits

Deadline: 2022-06-06

Location: Reading, UK or Bonn, Germany

Notes: 4 year limited contract, with possibility of further contracts, starting ASAP

Side Reading:

• ECMWF Machine Learning Roadmap
• ECMWF 10-year strategy
• Destination Earth programme