Scholar
Lechao Xiao
Google Scholar ID: fvwzUnIAAAAJ
Google DeepMind
Citations & Impact
Citations
4,065
H-index
19
i10-index
24
Publications
20
Co-authors
13
list available
Contact
No contact links provided.
Publications
Resume (English only)
Academic Achievements
- - Publication: Rethinking Conventional Wisdom in Machine Learning: from Generalization to Scaling
- - Publication: Scaling Exponents Across Parameterizations and Optimizers
- - Publication: 4+3 Phases of Compute-Optimal Neural Scaling Laws
- - Publication: Small-scale proxies for large-scale Transformer training instabilities
- - Publication: Synergy and symmetry in deep learning: Interactions between the data, model, and inference algorithm
- - Publication: Precise Learning Curves and Higher-Order Scalings for Dot-product Kernel Regression
- - Publication: Eigenspace Restructuring: a Principle of Space and Frequency in Neural Networks
- - Publication: Fast Neural Kernel Embeddings for General Activations
- - Publication: Dataset Distillation with Infinitely Wide Convolutional Networks
- - Publication: Exploring the Uncertainty Properties of Neural Networks' Implicit Priors in the Infinite-Width Limit
- - Publication: Finite Versus Infinite Neural Networks: an Empirical Study
- - Publication: The Surprising Simplicity of the Early-Time Learning Dynamics of Neural Networks
- - Publication: Provable Benefit of Orthogonal Initialization in Optimizing Deep Linear Networks
- - Publication: Disentangling Trainability and Generalization in Deep Neural Networks
- - Publication: Neural Tangents: Fast and Easy Infinite Neural Networks in Python
- - Publication: Wide Neural Networts of Any Depth Evolve As Linear Models Under Gradients Descent
- - Publication: Bayesian Deep Convolutional Networks with Many Channels are Gaussian Processes
- - Publication: Dynamical isometry and a mean field theory of CNNs: How to train 10,000-layer vanilla convolutional neural networks
Research Experience
- - Research Scientist: Google DeepMind (legacy Google Brain), NYC
- - Hans Rademacher Instructor of Mathematics: University of Pennsylvania
Education
- - PhD: University of Illinois at Urbana-Champaign
- - BA: Zhejiang University, Hangzhou, China
Background
- Research interests include scaling-centric machine learning, deep learning theory, generalization, optimization, training dynamics, kernels, Gaussian processes, etc. In his previous research, he also worked on harmonic analysis.
