Mert Pilanci
Associate Professor
Develops convex optimization formulations of neural network training and randomized methods for large-scale learning.
Mert Pilanci's lab at Stanford works on learning theory & optimization. 58 papers since 2023, 10 at top venues, mostly ICLR and ICML. 8 PhD students. 9 alumni. Applications close .
At a glance
- papers, 2023–2026
- 15/ year
- at T1 venues
- 17%
- PhD students
- 8
- MS students
- 0
- alumni listed
- 9
Papers by yearfrom the lab homepage, 2023–2026
- 16 papers2026 · T1 6 · collecting
- 10 papers2025 · T1 3
- 16 papers2024 · T1 0
- 16 papers2023 · T1 1
T1 venuesOtherVenues: ICLR, ICML, TASLP, TMLR, ICASSP
Research focuspaper titles that use each topic's words
- Learning theory & optimization18 papers
- convex16 papers
- convex optimization5 papers
- randomized4 papers
- network training0 papers
Paper list
From the lab homepage.
202616 papers
- SIAM Journal on Mathematics of Data ScienceA Library of Mirrors: Deep Neural Nets in Low Dimensions are Convex Lasso Models with Reflection Features (opens in a new tab)
- ICLR (T1)A Recovery Guarantee for Sparse Neural Networks (opens in a new tab)
- accepted to NeuripsAccelerating the Inference Era with AI-Driven, Globally Optimized HW/SW Co-Design
- ICML AI4Math WorkshopAI-Assisted Discovery of Convex Relaxations via Dual Agents (opens in a new tab)
- IEEE International Symposium on Information Theory (ISIT)An Information-Theoretic Perspective on LLM Tokenizers (opens in a new tab)
- ICLR (T1)ASMIL: Attention-Stabilized Multiple Instance Learning for Whole-Slide Imaging (opens in a new tab)
- accepted to IEEE Transactions on Information TheoryBlack Boxes and Looking Glasses: Multilevel Symmetries, Reflection Planes, and Convex Optimization in Deep Networks (opens in a new tab)
- ICLR (T1)CL-DPS: A Contrastive Learning Approach to Blind Nonlinear Inverse Problem Solving via Diffusion Posterior Sampling (opens in a new tab)
- ICML (T1)Convex Low-resource Accent-Robust Language Detection in Speech Recognition (opens in a new tab)
- ICML (T1)Convex Optimization for Alignment and Preference Learning on a Single GPU (opens in a new tab)
- TMLR (T1)Deep Neural Nets in Low Dimensions with Sign Activations are Convex Lasso Models (opens in a new tab)
- —FLASHSKETCH: Sketch-Kernel Co-Design for Fast Sparse Sketching on GPUs (opens in a new tab)
202510 papers
- International Conference on Machine Learning (ICML)Active Learning of Deep Neural Networks via Gradient-Free Cutting Planes (opens in a new tab)
- TASLP (T1)Disentangling Speech Representations Learning With Latent Diffusion for Speaker Verification (opens in a new tab)
- International Conference on Learning Representations (ICLR) - Oral PresentationExploring the Loss Landscape of Regularized Neural Networks via Convex Duality (opens in a new tab)
- International Conference on Machine Learning (ICML)Geometric Algebra Planes: Convex Implicit Neural Volumes (opens in a new tab)
- TASLP (T1)Mutual Information-Enhanced Contrastive Learning With Margin for Maximal Speaker Separability (opens in a new tab)
- International Conference on Learning Representations (ICLR)Newton Meets Marchenko-Pastur: Massively Parallel Second-Order Optimization with Hessian Sketching and Debiasing (opens in a new tab)
- IEEE Transactions on Information TheoryOverparameterized ReLU Neural Networks Learn the Simplest Model: Neural Isometry and Phase Transitions (opens in a new tab)
- IEEE Conference on Decision and Control (CDC)Scalable Power Management of Data Centers via Proximal Gradient Optimization (opens in a new tab)
- IEEE Transactions on Information TheoryThe Convex Landscape of Neural Networks: Characterizing Global Optima and Stationary Points via Lasso Models (opens in a new tab)
- ICASSP (T1)Thinking While Listening: Simple Test Time Scaling For Audio Classification (opens in a new tab)
202416 papers
- International Conference on Acoustics, Speech, and Signal Processing (ICASSP)Adaptive Large Language Models By Layerwise Attention Shortcuts (opens in a new tab)
- Neural Information Processing Systems (Neurips)Adaptive Sampling for Efficient Softmax Approximation
- Neural Information Processing Systems (Neurips)Compressing Large Language Models using Low Rank and Low Precision Decomposition (opens in a new tab)
- International Conference on Machine Learning (ICML)Convex Relaxations of ReLU Neural Networks Approximate Global Optima in Polynomial Time (opens in a new tab)
- International Conference on Acoustics, Speech, and Signal Processing (ICASSP)ConvexECG: Lightweight and Explainable Neural Networks for Personalized, Continuous Cardiac Monitoring (opens in a new tab)
- Neural Information Processing Systems (Neurips)CRONOS: Enhancing Deep Learning with Scalable GPU Accelerated Convex Neural Networks (opens in a new tab)
- IEEE Transactions on Information TheoryGradient Coding through Iterative Block Leverage Score Sampling (opens in a new tab)
- Mathematical ProgrammingNeural Spectrahedra and Semidefinite Lifts: Global Convex Optimization of Polynomial Activation Neural Networks in Polynomial-Time (opens in a new tab)
- SIAM Journal on Mathematics of Data ScienceOptimal Neural Network Approximation of Wasserstein Gradient Direction via Convex Optimization (opens in a new tab)
- ICCPower-Managed Data Centers for Sustainable Computing (opens in a new tab)
- Journal of the Royal Statistical Society, Series BPretraining and the Lasso (opens in a new tab)
- International Conference on Machine Learning (ICML)Riemannian Preconditioned LoRA for Fine-Tuning Foundation Models (opens in a new tab)
202316 papers
- Wiley Magnetic Resonance in MedicineCoil Sketching for Computationally-Efficient MR Iterative Reconstruction (opens in a new tab)
- International Conference on Acoustics, Speech, and Signal Processing (ICASSP)Convex Optimization of Deep Polynomial and ReLU Activation Neural Networks (opens in a new tab)
- IEEE Transactions on Information TheoryDistributed Sketching for Randomized Optimization: Exact Characterization, Concentration and Lower Bounds (opens in a new tab)
- Neural Information Processing Systems (NeurIPS)Fixing the NTK: From Neural Network Linearizations to Exact Convex Programs (opens in a new tab)
- TMLR (T1)From Complexity to Clarity: Analytical Expressions of Deep Neural Network Weights via Clifford's Geometric Algebra and Convexity (opens in a new tab)
- International Conference on Learning Representations (ICLR)Globally Optimal Training of Neural Networks with Threshold Activation Functions (opens in a new tab)
- International Conference on Acoustics, Speech, and Signal Processing (ICASSP)Low Precision Representations for High Dimensional Models (opens in a new tab)
- Journal of Linear Algebra and Its ApplicationsM-IHS: An Accelerated Randomized Preconditioning Method Avoiding Costly Matrix Decompositions (opens in a new tab)
- Neural Information Processing Systems (NeurIPS)Matrix Compression via Randomized Low Rank and Low Precision Factorization (opens in a new tab)
- International Conference on Machine Learning (ICML)Optimal Sets and Solution Paths of ReLU Networks (opens in a new tab)
- International Conference on Machine Learning (ICML)Optimal Shrinkage for Distributed Second-Order Optimization (opens in a new tab)
- International Conference on Learning Representations (ICLR)Parallel Deep Neural Networks Have Zero Duality Gap (opens in a new tab)
Where alumni went9 listed
Apply to StanfordFall 2027
Stanford Computer Science PhD
- Round
- Single round
- Opens
- GRE
- See program page
- English
- TOEFL/IELTS
- Fee
- See program page
Confirmed on the program page.
Similar labsnearest by research summary
- USUMD
William Regli
- Interpretability, evaluation & safety of LMs
- USUT Austin
Sujay Sanghavi
- Learning theory & optimization
- USUCSD
Tianhao Wang
- Learning theory & optimization
- CHETH Zurich
Florian Tramèr
- Trustworthy & safe ML
- USGeorgia Tech
Mayya Zhilova
- Learning theory & optimization
- CHEPFL
Nicolas Flammarion
- Learning theory & optimization
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
Stanfordcloses