Understanding-based research in Machine and Deep Learning.
10 Apr 2026, The fourth workshop on Metascience for Machine Learning.
Our fourth workshop is on Friday, April 10th, 2026 from 10:00-15:00; tentative schedule, and location is Lecture room Pi at Building 36, TU Delft campus, Mekelweg 4; 2628 CD Delft; register here.
Schedule
10:00: Walk in with coffee/tea
10:15 Overview presentation
10:30 workgroup updates and new workgroup stimulation.
11:00: International workshop pitch by Jan van Gemert (Delft University of University), about “E.T.: Empirical Theory” for stimulating theory-driven empirical research in Representation Learning.
12:00: Lunch
13:00: Talk by Hannah Pinson (Eindhoven University of Technology), about “A research recipe for understanding-focused deep learning research”.
14:00: Discussion.
15:00: End
Speaker presentations:
Pitch by Jan van Gemert (Delft University of University).
Title: “E.T.: Empirical Theory” for stimulating theory-driven empirical research in Representation Learning through Deep Learning
Abstract: .
Invited talk by Hannah Pinson (Eindhoven University of Technology)
Title: A research recipe for understanding-focused deep learning research
Abstract: Neural networks are not the black boxes a lot of people think they are. However, uncovering their inner workings is no small feat, and a clear methodology to obtain deeper insights in how they work seems lacking. In this talk, I discuss my own ‘understanding-focused deep learning research recipe’ , using the results obtained in the (work-in-progress) paper “It’s not a Lottery, It’s a Race: Understanding How Gradient Descent Adapts the Network’s Capacity to the Task” as an example project. I will also talk about the challenges I encountered, and share some practical insights, which I hope will serve as a starting point for a broader discussion on this exciting topic. .
