Radu Marculescu
Professor
Develops ML methods for embedded and edge systems, including efficient vision transformers, graph neural networks and federated learning.
Radu Marculescu's lab at UT Austin works on trustworthy & safe ML, graph learning and efficient training & inference. 56 papers since 2023, 17 at top venues, mostly WACV and CVPR. 5 PhD students. 47 alumni, most often to ARM ML Technology and Assistant Professor Univ. of Southern California Los Angeles CA Dept. of EE-Systems.. Applications close .
At a glance
- papers, 2023–2026
- 15/ year
- at T1 venues
- 30%
- PhD students
- 5
- MS students
- 0
- alumni listed
- 47
Papers by yearfrom the lab homepage, 2023–2026
- 9 papers2026 · T1 4 · collecting
- 17 papers2025 · T1 5
- 15 papers2024 · T1 4
- 15 papers2023 · T1 4
T1 venuesOtherVenues: WACV, CVPR, ICLR, NeurIPS, AAAI
Research focuspaper titles that use each topic's words
- Graph learning20 papers
- Efficient training & inference15 papers
- Trustworthy & safe ML1 paper
- transformers0 papers
- embedded0 papers
Paper list
From the lab homepage.
20269 papers
- WACVAdaptViG: Adaptive Vision GNN with Exponential Decay Gating
- The Fifth Learning on Graphs ConferenceAttentionViG: Cross-Attention-Based Dynamic Neighbor Aggregation in Vision GNNs
- IEEE Transactions on Computer-Aided Design of Integrated Circuits and SystemsBHAR: A System-Level Optimization Framework for Multisensor Batteryless Human Activity Recognition
- CVPR (T1)Fuel Gauge: Estimating Chain-of-Thought Length Ahead of Time in Large Multimodal Models
- AAAI (T1)NeuS-QA: Grounding Long-Form Video Understanding in Temporal Logic and Neuro-Symbolic Reasoning
- CVPR WorkshopObjectAlign: Neuro-Symbolic Object Consistency Verification and Correction
- CVPR (T1)Single-Round Scalable Analytic Federated Learning
- WACVSmoothDiffusion-VE: Real-time Generative Video Editing Using Adaptive Feature Cache
- NeurIPS (T1)Video Compression Meets Video Generation: Latent Inter-Frame Pruning with Attention Recovery
202517 papers
- WACVAda-VE: Training-Free Consistent Video Editing Using Adaptive Motion Prior
- IEEE International Conference on Body Sensor NetworksBatteryless Gesture Recognition Via Learned Sampling
- SenSys (T1)Demo Abstract: Lightweight Training and Inference for Self-Supervised Depth Estimation on Edge Devices
- MICCAI (T1)EfficientMedNeXt: Multi-Receptive Dilated Convolutions for Medical Image Segmentation
- CVPR (T1)EffiDec3D: An Optimized Decoder for High-Performance and Efficient 3D Medical Image Segmentation
- DCOSS-IoTFederated Continual Learning for Monocular Depth Estimation in Dynamic Indoor Environments
- Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) WorkshopsFrom Data to Design: Leveraging Frequency Statistics for Efficient Neural Network Architectures
- Advances in Social Networks Analysis and Mining, Springer, , ISBN: 978-3-031-85386-9Graph Learning for Bidirectional Disease Contact Tracing on Real Human Mobility Data
- CAIHealth Status Discovery for Online Bidirectional Contact Tracing and Disease Aware Navigation
- NeurIPS (T1)LoMix: Learnable Weighted Multi-Scale Logits Mixing for Medical Image Segmentation
- NeurIPS (T1)Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models
- Proceedings of the International Conference on Computer Vision (ICCV) WorkshopsMK-UNet: Multi-kernel Lightweight CNN for Medical Image Segmentation
202415 papers
- Proceedings of the ACM/IEEE Conference on Internet of Things Design and ImplementationBeyond Thresholds: A General Approach to Sensor Selection for Practical Deep Learning-based HAR
- Proceedings of the ACM/IEEE Conference on Internet of Things Design and ImplementationCHESSFL: Clustering Hierarchical Embeddings for Semi-Supervised Federated Learning
- Proceedings of the ACM/IEEE Conference on Internet of Things Design and ImplementationDemo Abstract: Online Training and Inference for On-Device Monocular Depth Estimation
- CVPR (T1)EMCAD: Efficient Multi-scale Convolutional Attention Decoding for Medical Image Segmentation
- —Fairness Implications of Machine Unlearning: Bias Risks in Removing NSFW Content from Text-to-Image Models
- CVPR WorkshopFast-NTK: Parameter-Efficient Unlearning for Large-Scale Models
- WACVG-CASCADE: Efficient Cascaded Graph Convolutional Decoding for 2D Medical Image Segmentation
- CVPR (T1)GreedyViG: Dynamic Axial Graph Construction for Efficient Vision GNNs
- ICLR (T1)Machine Unlearning for Image-to-Image Generative Models
- LOGMulti-Scale High-Resolution Logarithmic Grapher Module for Efficient Vision GNNs
- BSNPacket Pruning: Finding Better Energy Spending Policies for Batteryless Human Activity Recognition
- Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) WorkshopsPP-SAM: Perturbed Prompts for Robust Adaption of Segment Anything Model for Polyp Segmentation
202315 papers
- IEEE Transactions on Emerging Topics in ComputingCANNON: Communication-Aware Sparse Neural Network Optimization
- IoTDIDemo Abstract: A Hardware Prototype Targeting Federated Learning with User Mobility and Device Heterogeneity
- Multimodal Learning and Applications WorkshopDynamic Multimodal Fusion
- Thirty- Conference on Neural Information Processing Systems (NeurIPS)Efficient Low-rank Backpropagation for Vision Transformer Adaptation
- CVPR (T1)Efficient On-device Training via Gradient Filtering
- WACVMedical Image Segmentation via Cascaded Attention Decoding
- Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) WorkshopsMobileViG: Graph-Based Sparse Attention for Mobile Vision Applications
- Proceedings of the ACM/IEEE Conference on Internet of Things Design and ImplementationMOHAWK: Mobility and Heterogeneity-Aware Dynamic Community Selection for Hierarchical Federated Learning
- Medical Imaging with Deep LearningMulti-scale Hierarchical Vision Transformer with Cascaded Attention Decoding for Medical Image Segmentation
- ASONAMQuarantine in Motion: A Graph Learning and Multi-Agent Reinforcement Learning Framework to Reduce Disease Transmission Without Lockdown
- ICLR (T1)Revisiting Pruning at Initialization through the Lens of Ramanujan Graph
- IEEE Transactions on Computers, , ISSN: 0018SUGAR: Efficient Subgraph-level Training via Resource-aware Graph Partitioning
Where alumni went47 listed, 75% to industry
Industry 6Academia 2
Most often to
- ARM ML Technology
- Assistant Professor Univ. of Southern California Los Angeles CA Dept. of EE-Systems.
- Intel
Also to
- Texas Tech University
- Tsinghua University
Apply to UT AustinFall 2027
Computer Science PhD
- Round
- Single round (Dec 15)
- Opens
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- SGNUS
Wee Sun Lee
- Probabilistic & causal inference
- CNZhejiang
Yingming Li
- Data management & big data
- CAToronto
Scott Sanner
- Reinforcement learning
- USUT Austin
Haris Vikalo
- Trustworthy & safe ML
- USNYU
Andrew Gordon Wilson
- Probabilistic & causal inference
- CNTsinghua
Yanyan Lan
- Recommendation, search & web
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.
UT Austincloses