Reverse-Engineering an Azure AutoML Forecasting Model
Hi friends,
I was wondering - do you use AutoML in your day to day life? And not just the Azure flavor but, like, any of them?
Me, I find AutoML useful for training an initial baseline model. You know, when you've just managed to get a dataset that's just clean enough to be useful, and you're itching to see if this time you'll need to use something other than LightGBM. It's always LightGBM though. Always. It's that good.
Anyway, I don't really use AutoML apart from this initial phase. It's slow enough to make it impractical for regular retrains, and when you need a model that's constantly updated, well, slow is out and fast is in.
What I do instead is reverse engineer the AutoML model, take whatever I can from there and adapt whatever I can't take. Kinda like Picasso, I imagine 🙂.
If you're curious about the process, I've written a short(er) blog post on creating your own models based on Azure AutoML-trained ones. You'll find it here. It's focused on time series, but the underlying principles should apply to regression and classification models, too.
Hope you like it. And I appreciate any and all feedback you may have - just reply directly to this email, or head over to Twitter and let me know.
Thanks for reading,
Vlad
Vlad Iliescu
Click here to read or share this on the web. Hi friends, This was a lighter month in terms of writing articles, but don't fret - I've still got some interesting things for you to read. For starters, if you want to understand more about craftsmanship and about honing your machine learning skills, you might want to read this thread 🧶, the distillation of my conversation with Sandro Mancuso from earlier this month. I've also a couple of tips on Azure Functions, stuff like how to configure...
Hi friends, We all know that machine learning projects are quite different from standard software development projects. For example, estimating an ML project is surprisingly challenging, to the point of being impossible at times. Maintaining an ML project, specifically deciding when to update a model or a pipeline or a dataset is quite different too when compared to maintaining a standard dev project. This of course comes as no surprise, since developing a classic system requires code, code,...
Hi friends, The future is coming and there's reason to be excited about it. I've been lucky enough to be able to test GitHub Copilot for the past few weeks and even though it has its quirks, it looks very promising. Not promising enough to fear losing our jobs, but promising enough to warrant daydreaming about the day when we'll all be using Copilot-powered compilers to compile English to Python/C#/JavaScript/etc. Sigh. I've even written an article about my experience. It has significantly...