The Curious Engineer

All blogs / Math

Machine Learning Research Blog

francisbach.com

Francis Bach

MathAI / MLAlgorithms / data structures

’19’20’21’22’23’24’25’26
Posts
42 from 27 Jun 2019 to 25 Sep 2026
Posting
now and then (2 in the last year)
Kind
personal blog
Feed
https://francisbach.com/feed/

2026 2

  1. Exploding variance of means of exponentials: least-squares to the rescue
  2. Closed-form dynamics beyond quadratics

2025 3

  1. Beyond Power Laws: Scaling Laws for Next-Token Prediction
  2. Revisiting scaling laws via the z-transform
  3. Unraveling spectral properties of kernel matrices – II

2024 3

  1. My book is (at last) out!
  2. Scaling laws of optimization
  3. Unraveling spectral properties of kernel matrices – I

2023 2

  1. Revisiting the classics: Jensen’s inequality
  2. Non-convex quadratic optimization problems

2022 7

  1. Discrete, continuous and continuized accelerations
  2. Sums-of-squares for dummies: a view from the Fourier domain
  3. Rethinking SGD’s noise – II: Implicit Bias
  4. Rethinking SGD’s noise
  5. Information theory with kernel methods
  6. Playing with positive definite matrices – II: entropy edition
  7. Playing with positive definite matrices – I: matrix monotony and convexity

2021 6

  1. Approximating integrals with Laplace’s method
  2. The quest for adaptivity
  3. I am writing a book!
  4. Going beyond least-squares – II : Self-concordant analysis for logistic regression
  5. Going beyond least-squares – I : self-concordant analysis of Newton method
  6. Finding global minima with kernel approximations

2020 12

  1. Optimization is as hard as approximation
  2. The Cauchy residue trick: spectral analysis made “easy”
  3. Polynomial magic III : Hermite polynomials
  4. The many faces of integration by parts – II : Randomized smoothing and score functions
  5. The many faces of integration by parts – I : Abel transformation
  6. Gradient descent for wide two-layer neural networks – II: Generalization and implicit bias
  7. Gradient descent for wide two-layer neural networks – I : Global convergence
  8. Effortless optimization through gradient flows
  9. Computer-aided analyses in optimization
  10. On the unreasonable effectiveness of Richardson extrapolation
  11. Acceleration without pain
  12. The sum of a geometric series is all you need!

2019 7

  1. Polynomial magic II : Jacobi polynomials
  2. Polynomial magic I : Chebyshev polynomials
  3. Are all kernels cursed?
  4. The Gumbel trick
  5. The “η-trick” reloaded: multiple kernel learning
  6. The “η-trick” or the effectiveness of reweighted least-squares
  7. I am starting a blog!