differentiable SDE machine learning is so confusing

[sw1ftp@w]

Can anyone tell me what I should do to understand differentiable SDE machine learning more easily?

I feel like SDE or Stochastic Differential Equations is so hard and complex. I can't even find some additional resources

Does anyone here have info about SDE, like blogs, books, papers, or even youtube videos that can help me understand SDE easier?

 

[Bruynee]

How are you approaching SDEs right now, are you starting with the math foundations or jumping straight into the ML side? It might help to build up from basic Ito calculus before diving into ML papers.

 

[desperate-ice466]

You can look up blogs from Ludwig Winkler, it breaks down Itô's Lemma, solving SDEs, and more in a beginner-friendly way. Also, the book An Introduction to Stochastic Differential Equations by L.C. Evans is a good resource if you're ready to dive deeper.

 

[TerMinalist3]

A good way to start is by building a solid grasp of basic stochastic calculus, think Itô's lemma and Brownian motion, stuff like that then from there, look for machine learning papers that break down differentiable SDEs with clear examples, YouTube channels and blogs explain SDEs in bite sized pieces, mix intuition and maths

 

[Parker]

If SDEs feel impossible, maybe the problem isn't the math, it's that machine learning folks jumped on a hype train they don't fully grasp. Start with calculus and probability first, don't let academia gaslight you

 

[josef.gml11]

Can anyone tell me what I should do to understand differentiable SDE machine learning more easily?

I feel like SDE or Stochastic Differential Equations is so hard and complex. I can't even find some additional resources

Does anyone here have info about SDE, like blogs, books, papers, or even youtube videos that can help me understand SDE easier?

SDEs are challenging because they combine calculus with probability. So the key is to build your understanding step-by-step, starting with core concepts before you dive into the complex math.

I agree with @TerMinalist3 that it's crucial to build a solid grasp of Brownian motion, which is the stochastic part of SDEs. Once you have an intuitive feel for that, you'll be in a much better position to understand Itô calculus and the rest of the theory. Try checking online for blogs and videos that focus on intuition over rigorous proofs. Then when you're ready for a deeper dive, check out the book "Applied Stochastic Differential Equations" by Särkkä and Solin, as it's specifically geared toward machine learning.

Just keep at it and take your time to understand each piece, and it will eventually click!

 

[Bob2000]

I faced the same problem, for me when reading papers mention these SDEs. I just read their equation and the proof, then googled the wikipedia pages for the original equations for example Heun or Euler method and tried to understand their work.

I tried to dig deeper to how they prove it but I gave up after spending hours and couldn't understand anything