Simon Prince

Professor of Computer Science, University of Bath

[Simon Prince @SimonPrinceAI]

Are you (or is someone you know) teaching AI / ML / Deep Learning this year? My forthcoming book (freely available at https://t.co/hqRA1xUPkk) will save you a lot of time. This thread will show you why.

[Simon Prince @SimonPrinceAI]

@ilkerarslan_35 Don't worry... it's not that different and it only goes to chapter eight anyway. I'm honestly very grateful that you have ordered the hard copy. Thanks!

[Simon Prince @SimonPrinceAI]

I'm gradually updating and improving chapters from Understanding Deep Learning and posting them to https://t.co/8zGsFiXjXd.

The latest update concerns how to measure the performance of the trained models.

[Simon Prince @SimonPrinceAI]

@lfxy9 It has already been translated to Chinese. I own a copy. You should be able to buy it online.

[Simon Prince @SimonPrinceAI]

I've always really hated this illustration of byte-pair encoding from my book, but can't figure out how to make it better. Anyone out there got some design skills and want to have a crack at it? DM me if you are interested.

[Simon Prince @SimonPrinceAI]

My latest tutorial for @RBCBorealis considers Neural ODEs. These extend residual networks to have an infinite number of layers, each of which makes an infinitesimal change to the data representation and underpin generative models like flow-matching.

📙 https://t.co/x6zFx4G1q1

[Simon Prince @SimonPrinceAI]

Understanding Deep Learning, now published in Japanese (also available in Russian, Chinese and Korean).

[マーキュリー２世 @uranus_2]

・Prince: Understanding Deep Learning (MIT Press, 2023)を2分冊で全訳.オールカラー
・世界標準の教科書
・ChatGPTなど生成AIのしくみを理解するために.
・I巻では基礎となる教師あり学習からはじめ,トランスフォー…⇒深層学習を理解する I: 教師あり学習とモデルの基礎　https://t.co/Eo640bZja0

[Simon Prince @SimonPrinceAI]

@blackpillah I'd say start with iclimbtrees until it runs out of material (chapter 8) and then switch to the book!

[Simon Prince @SimonPrinceAI]

I'm updating and improving chapters from Understanding Deep Learning and posting them to https://t.co/8zGsFiXjXd.

The latest unit describes a framework for loss functions from which least squares, binary cross-entropy, and multiclass cross-entropy losses can be derived.

[Simon Prince @SimonPrinceAI]

@2023gianni Bad decision I made 16 years ago, but trying to be internally consistent!

[Simon Prince @SimonPrinceAI]

I'm gradually updating and improving chapters from Understanding Deep Learning and posting them to https://t.co/8zGsFiXjXd .

The latest update concerns the backpropagation algorithm and network initialization. Critical info if you are teaching yourself ML!

[Simon Prince @SimonPrinceAI]

@Ananth2394 I'd probably read the site first, until the content runs out and then switch to the book.

[Simon Prince @SimonPrinceAI]

@Proizve Fixed! Much appreciated that you let me know about this.

[Simon Prince @SimonPrinceAI]

I'm gradually updating and improving chapters from Understanding Deep Learning and posting them to https://t.co/8zGsFiXjXd .

The latest addition considers how to find the parameters that minimize the loss. This process is known as training the model.

[Simon Prince @SimonPrinceAI]

@Proizve Thanks so much. I'm looking into it. Nothing has been updated since July so it's a bit mysterious.

[Simon Prince @SimonPrinceAI]

@GiorgioMantova 2. I use the full ∂L/∂φ notation just because it's the most well known and neutral. All sorts of people read my book... psychologist, economists, materials scientists, so just going for what I think will be best understood by everyone.

[Simon Prince @SimonPrinceAI]

@GiorgioMantova 1. I use square brackets for functions and round for probability distributions. It's not standard and was intended to distinguish from "grouping brackets". I made this decision for my first book in 2012 and it was probably a mistake. Now just aiming for internal consistency.

[Simon Prince @SimonPrinceAI]

Ever wondered why larger learning rates and smaller batches lead to better generalization?

Part XI of the ODE/SDE in ML series for @RBCBorealis, reaveals why by treating stochastic gradient descent in the infinitesimal limit as an SDE.

https://t.co/4BMPSD6FmS

[Simon Prince @SimonPrinceAI]

Over the last year, I've written 8 blogs for @RBCBorealis that will teach you everything you need to know about ODEs and SDEs for machine learning:

https://t.co/r7J2k6j6DT

Coming soon: neural ODEs, continuous normalizing flows, diffusion models, flow matching, PINNs and more.

[Simon Prince @SimonPrinceAI]

@afrenkai @RBCBorealis I'm going to compile all of these into a separate book that should come out next year.

[Simon Prince @SimonPrinceAI]

Part 8 of my series on ODEs and SDEs in machine learning for @RBCBorealis introduces the Fokker-Planck equation and Anderson's reverse time theorem. The former maps an SDE to a PDE over a probability distribution. The latter reverses an SDE in time.

https://t.co/h7x44cOu9U

[Kosta Derpanis @CSProfKGD]

I’ve launched 2 new course websites for my Introduction to #ComputerVision and #DeepLearning in Computer Vision courses.

🆕 Gaussian Splatting
🆕 Flow Matching
🆕 Sparse Autoencoders
🆕 Lecture Notes (work in progress)

The videos don’t include voiceovers, but that’s coming in a future revision 🤞

Share your feedback in the thread.