⭐ In navigation GMaps gets an A and a star
Late to the Party 🎉 is about insights into real-world AI without the hype.
Hello internet,
first vacationing, then a bad cold had me miss a couple of issues, but we’re back in full force with some lovely machine learning!
The Latest Fashion
- It’s official! The first alpha of our ML-based weather forecasting system at ECMWF is live!
- You can now use Dall-E 3 through Bing (even without Edge!)
- Kaggle published their AI report for 2023
Got this from a friend? Subscribe here!
My Current Obsession
Right after returning from my travels, I got that cold that was going around. But now I’m back! Lots to catch up on!
Thing I Like
This week, I put up a bunch of autumny and Halloweeny decorations, one of which is a witches cauldron with one of those ultrasonic fog makers in it to make it fun and bubbling! Such a small and fun idea to honstely. I just hope I don’t end up hearing it because bats usually drive me crazy when they’re around… Wish me luck!
Hot off the Press
I haven’t really been writing anything, but hopefully, I’ll get back in the groove soon.
On Socials
Announcing the ECMWF AIFS data-driven weather forecasting system has gone over pretty well, also on Mastodon!
Linkedin also really enjoyed the Learning Interpretability Tool to inspect LLMs, which I shared in this newsletter 6 months ago, as well as, the ML Engineering flashcards from 13 months ago! Mastodon enjoyed this GPS Jamming map from a year ago.
Python Deadlines
The PyTexas 2024 conference proposal deadlines have been announced!
Data Stories
How does Google Maps find the best way to your goal?
This YouTube video illustrates the A* search algorithm.
This is a graph-based algorithm that interprets maps as graphs, which I find personally is a natural interpretation.
Imagine you're in Rome, and you want to find the best way to explore the city's famous landmarks like the Colosseum, the Roman Forum, and the Vatican. A* will be your trusty guide.
Here's how A* works for your Roman adventure:
Starting Point and Destination: You start your day at your home and plan to end at your friend's house across town. These are your start and end points, like your travel itinerary.
Possible Routes: A* looks at the map of Rome, with all the streets, squares, and attractions. It considers different paths and streets you could take.
Cost and Heuristic: A* calculates not only the distance between these attractions but also considers a heuristic, which is like an estimate of how far it is "as the crow flies" between these points. This helps in making educated guesses about how much more exploring is left. So, it combines the actual distance you've covered with this estimation. This is important to enable exploration but keep the algorithm fast and focused.
Priority Queue: A* keeps a list of the streets you could visit, along with the estimated cost of getting from one to another. It uses a priority queue to decide which street to enter next. The one with the lowest combined cost (actual distance + heuristic) goes to the top of your list to try.
Exploring Paths: A* starts at your location and looks at nearby directions. It chooses the one with the lowest estimated cost to visit next.
Repeating the Process: A* repeats this process as you journey through Rome. It keeps track of the best path, updating it whenever it finds a more efficient route to your next destination.
Optimal Path: Finally, A* guides you along the optimal path to reach the destination. Now if we wanted to take traffic and travel-time into account, we could actually do this! We just need to adjust our graph from distance-based to travel-time-based!
And who knew it could look so beautiful!
Source: Youtube
Question of the Week
- How do you treat missing values in your data?
Post them on Mastodon and Tag me. I'd love to see what you come up with. Then, I can include them in the next issue!
Tidbits from the Web
- This Tiktok caught a lightning strike in an active volcano
- Last week was Fat Bear Week, a celebration of bears chonking up before hibernation
- This skit about performance reviews had me cackling
Jesper Dramsch is the creator of PythonDeadlin.es, ML.recipes, data-science-gui.de and the Latent Space Community.
I laid out my ethics including my stance on sponsorships, in case you're interested!