Mahdi Soltanolkotabi
Andrew and Erna Viterbi Early Career Chair and Professor of Electrical and Computer Engineering, Computer Science, and Industrial and Systems Engineering
Researches theory of deep learning, optimization and AI foundations for science.
Mahdi Soltanolkotabi's lab at USC works on learning theory & optimization. 24 papers since 2023, 16 at top venues, mostly ICML and NeurIPS. 6 PhD students. 2 alumni, most often to Amazon. Applications close .
At a glance
- papers, 2023–2026
- 6/ year
- at T1 venues
- 67%
- PhD students
- 6
- MS students
- 0
- alumni listed
- 2
Papers by yearfrom DBLP, 2023–2026
- 4 papers2026 · T1 3 · collecting
- 11 papers2025 · T1 5
- 4 papers2024 · T1 4
- 5 papers2023 · T1 4
T1 venuesOtherVenues: ICML, NeurIPS, ICLR, ACC, COLT
Research focuspaper titles that use each topic's words
- Learning theory & optimization8 papers
- optimization1 paper
- foundations0 papers
- deep learning0 papers
- theory0 papers
Paper list
From DBLP.
20264 papers
- ICLR (T1)FoNE: Precise Single-Token Number Embeddings via Fourier Features (opens in a new tab)
- IEEE Trans. Inf. TheoryGradient Descent Provably Solves Nonlinear Tomographic Reconstruction (opens in a new tab)
- ICLR (T1)Learning to Recall with Transformers Beyond Orthogonal Embeddings (opens in a new tab)
- COLT (T1)When Both Layers Learn: Training Dynamics of Representing Linear Models via ReLU Networks (opens in a new tab)
202511 papers
- ICBCCryptoMamba: Leveraging State Space Models for Accurate Bitcoin Price Prediction (opens in a new tab)
- NeurIPS (T1)Emergence and Evolution of Interpretable Concepts in Diffusion Models (opens in a new tab)
- CVPR WorkshopHARMONY: Hidden Activation Representations and Model Output-Aware Uncertainty Estimation for Vision-Language Models (opens in a new tab)
- NeurIPS (T1)Hyperphantasia: A Benchmark for Evaluating the Mental Visualization Capabilities of Multimodal LLMs (opens in a new tab)
- IEEE Trans. Inf. TheoryImplicit Balancing and Regularization: Generalization and Convergence Guarantees for Overparameterized Asymmetric Matrix Sensing (opens in a new tab)
- ICIPLearning from PU Data Using Disentangled Representations (opens in a new tab)
- ICLR (T1)MediConfusion: Can you trust your AI radiologist? Probing the reliability of multimodal medical foundation models (opens in a new tab)
- ACCOn the stability properties of gradient flow dynamics for a symmetric low-rank matrix factorization problem (opens in a new tab)
- ICML (T1)Test-Time Training Provably Improves Transformers as In-context Learners (opens in a new tab)
- NeurIPS (T1)The Rich and the Simple: On the Implicit Bias of Adam and SGD (opens in a new tab)
- Trans. Mach. Learn. Res.Theoretical Insights into Overparameterized Models in Multi-Task and Replay-Based Continual Learning (opens in a new tab)
20244 papers
- ICML (T1)Adapt and Diffuse: Sample-adaptive Reconstruction via Latent Diffusion Models (opens in a new tab)
- ICML (T1)DiracDiffusion: Denoising and Incremental Reconstruction with Assured Data-Consistency (opens in a new tab)
- JMLR (T1)Learning from many trajectories (opens in a new tab)
- ICML (T1)Provable Multi-Task Representation Learning by Two-Layer ReLU Neural Networks (opens in a new tab)
20235 papers
- NeurIPS (T1)A Data-Free Approach to Mitigate Catastrophic Forgetting in Federated Class Incremental Learning for Vision Tasks (opens in a new tab)
- CVPR (T1)CUDA: Convolution-Based Unlearnable Datasets (opens in a new tab)
- NeurIPS (T1)Learning Provably Robust Estimators for Inverse Problems via Jittering (opens in a new tab)
- Trans. Mach. Learn. Res.mL-BFGS: A Momentum-based L-BFGS for Distributed Large-scale Neural Network Optimization (opens in a new tab)
- ICML (T1)On the Role of Attention in Prompt-tuning (opens in a new tab)
Where alumni went2 listed, 100% to industry
Industry 2Academia 0
Most often to
- Amazon
Apply to USCFall 2027
Computer Science PhD (Viterbi)
- Round
- Single round (Dec 15)
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Confirmed on the program page.
Similar labsnearest by research summary
- GBEdinburgh
Fengxiang He
- Trustworthy & safe ML
- CNZhejiang
Ziqi Yang
- Trustworthy & safe ML
- USNYU
Jonathan Niles-Weed
- Learning theory & optimization
- USUSC
Paria Rashidinejad
- Reinforcement learning
- USNYU
Gautam Kamath
- Learning theory & optimization
- CNTsinghua
Shu-Tao Xia
- Trustworthy & safe ML
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- DBLP
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
USCcloses