Luca Carlone

I am proud to share that our paper, "Greedy Recursive Spectral Bisection for Modularity-Bound Hierarchical Divisive Community Detection," has been published in Statistics and Computing. Thanks to an agreement between Springer and FCT - Portugal, you can access the full article for free! Dive into our innovative approach for community detection and explore the detailed methodology and results. Check it out here: https://rdcu.be/dL8bM #StatisticsAndComputing #CommunityDetection #Research #OpenAccess #Springer #PPCIC #PPPRO #CEFETRJ #DataScience

30

With a small amount of architectural extensions, our recently accepted paper "M^3XU: Achieving High-Precision and Complex Matrix Multiplication with Low-Precision MXUs" can let low-precision AI/ML-oriented matrix processors capable of performing IEEE 754 32-bit floating point, 64-bit double-precision operations and even more, native support for complex number operations! Programmers do not need to worry about changing their code or new accuracy issues -- the resulting architecture generates exactly the same outcome as standard 32-bit/64-bit floating numbers, but at least doubled performance! M^3XU will be published in SC24 in November and we will release the framework later! This work is done together with my students Dongho Ha Yunan Zhang at UCR, my collaborators, Chen-Chien Kao (UMich), Won Woo Ro (Yonsei University) and Christopher Hughes (Intel).

98

The increasing low-quality & hostile review s in the machine learning community are really concerning.

21

📢 Excited to share recent research achievements with my PhD student Reza Vafaee Here are three of our latest papers: 1- Learning-based Sparse Sensing with Performance Guarantees in IEEE Transactions on Automatic Control A novel method for converting weighted sensor schedules to sparse ones while maintaining performance. Read more: https://lnkd.in/e4eD2QGF 2- Real-time Sensor Selection for Time-Varying Networks with Guaranteed Performance in Automatica Dynamic sensor selection for LTV systems with high performance and minimal sensors. Read more: https://lnkd.in/eCYTmb9c 3- Exploring Non-Submodular Scheduling for Large-Scale Sensor Networks in IEEE Control Systems Letters An effective greedy algorithm for non-submodular sensor scheduling. Read more: https://lnkd.in/etxfgJaS Proud of our progress in sparse sensing in complex networks! 🚀 #Research #SensorNetworks #ControlSystems #SparseSensing #IEEE #Automatica #PhDResearch

64

4 Comments

[PAPER UPDATE] Excited to announce that our paper "When In-Memory Computing Meets Spiking Neural Networks: From Applications to Circuits" has been published in #AppliedPhysicsReviews! 🔥Curious about the trade-offs that shape cutting-edge designs? How do neuro-inspired spiking neural networks (SNNs) algorithms and In-memory Computing (IMC) revolutionize the future of AI? This paper dives deep into their synergies, investigating the complex interactions between accuracy, energy, latency, and area. Major Contributions: - Comprehensive overview of SNNs + IMC applications & circuits powering neuromorphic computing. - Exploiting SNN temporal sparsity for analog IMC with massive PPA gains—featuring #SpikeSim, an SNN-specific IMC architecture. - Tackling NVM non-idealities in IMC crossbars with solutions like noise-aware batchnorm, ensuring robust SNN inference. - Challenging the myth: "More NVM precision = better," highlighting trade-offs in energy efficiency, and area. -Future-forward insights on online learning, emerging neuronal devices, and co-exploration to drive energy-efficient IMC-SNN in next-gen AI hardware. Authors: Abhishek Moitra Abhiroop Bhattacharjee Yuhang Li Youngeun Kim Paper Link: https://lnkd.in/eDtFavrg SpikeSim Code: https://lnkd.in/eBz5ZnRD #NeuromorphicComputing #InMemoryComputing #SpikingNeuralNetworks #APR #CoDesign Center for the Co-Design of Cognitive Systems

156

1 Comment

Here's a fun short blog post by my students Shvetank Prakash Emil Njor Colby Banbury Matthew Stewart, PhD on Tiny Machine Learning (#tinyML).

68

3 Comments

Are you facing hurdles in scaling up your computational tasks or fine-tuning deep learning models on distributed memory systems with GPU acceleration? Join us at JuliaCon 2024 for the Dagger.jl Workshop! Discover how Dagger's parallel task-based runtime can empower you to efficiently parallelize across threads, nodes, and GPUs. Secure your spot now! 🌟 JuliaCon 2024 #JuliaLang #JuliaCon2024 #ParallelComputing #Daggerjl #GPUComputing

11

4 papers accepted for presentation at ECCV 2024, to be held in Milan, Italy from September 29 to October 4, 2024. Thanks to postdocs, PhD students and our collaborators! 1- Xinyu Yang, Hossein Rahmani, Sue Black, Bryan Williams, "Weakly Supervised Co-training with Swapping Assignments for Semantic Segmentation". Code and paper are available at https://lnkd.in/eCXrAsDr. The proposed method, CoSA, is the first single-stage weakly supervised semantic segmentation approach to outperform all existing multi-stage methods including those with additional supervision, surpassing existing baselines by a substantial margin. 2- Feixiang Zhou, Bryan Williams, Hossein Rahmani, "Towards Adaptive Pseudo-label Learning for Semi-Supervised Temporal Action Localization". Code and paper will be available soon 3- Zhengbo Zhang, Li Xu, Duo Peng, Hossein Rahmani, Jun Liu, "Diff-Tracker: Text-to-Image Diffusion Models are Unsupervised Trackers". Code and paper will be available soon. 4- Xiaofei Hui, Qian Wu, Hossein Rahmani, Jun Liu, "Class-Agnostic Object Counting with Text-to-Image Diffusion Model". Code and paper will be available soon. #ECCV2024

127

2 Comments

🚀 Attention all software engineers! ACM has launched the first issue of Proceedings of the ACM on Software Engineering (PACMSE), a Gold Open Access journal dedicated to advancing software engineering research. PACMSE covers a wide range of topics, from design to deployment, aiming to improve software quality. Each issue of the journal comprises a curated selection of papers, which will be announced through publicized calls for papers. For more detail on the submission process, refer to intro in the inaugural issue: https://bit.ly/3YwVBti  📚 #SoftwareEngineering #OpenAccess #ACM

59

1 Comment

ICRA2024 papers from the Montreal Robotics and Embodied Learning (REAL) Lab: Open X-Embodiment: Robotic Learning Datasets and RT-X Models Sergey Levine, Chelsea Finn, Ken Goldberg, Lawrence Yunliang Chen, Gaurav Sukhatme, Shivin Dass, Lerrel Pinto, Yuke Zhu, Yifeng Zhu, Shuran Song, Oier Mees, Deepak Pathak, Hao-Shu Fang, Henrik Iskov Christensen, Mingyu Ding, Youngwoon Lee, Dorsa Sadigh, ..., Kirsty Ellis, Glen Berseth, ... 280 other authors from 44 institutions... In this work we collaborated with over 40 institutions around the world to combine our robotics datasets and train a large VLM model over everything we collected. The results (https://lnkd.in/gYjjqEgZ), speak for themselves. The performance across a large range of tasks receives a large boost. This work won a best paper award at ICRA 2024! -------------------- ConceptGraphs: Open-Vocabulary 3D Scene Graphs for Perception and Planning 2024 Qiao Gu*, Alihusein Kuwajerwala*, Sacha Morin*, Krishna Murthy Jatavallabhula*, Bipasha Sen, Aditya Agarwal, Corban Rivera, William Paul, Kirsty Ellis, Rama Chellappa, Chuang Gan, Celso Miguel de Melo, Joshua B. Tenenbaum, Antonio Torralba, Florian Shkurti, Liam Paull Link: https://lnkd.in/g8gQ6m7x In this work, we propose ConceptGraphs, an open-vocabulary graph-structured representation for 3D scenes. ConceptGraphs is built by leveraging 2D foundation models and fusing their output to 3D by multi-view association. The resulting representations generalize to novel semantic classes, without the need to collect large 3D datasets or finetune models. We demonstrate the utility of this representation through a number of downstream planning tasks that are specified through abstract (language) prompts and require complex reasoning over spatial and semantic concepts. Social media posts: https://lnkd.in/gSAauXVi https://lnkd.in/gkbDkBkn ----------------------- Torque-based Deep Reinforcement Learning for Task-and-Robot Agnostic Learning on Bipedal Robots Using Sim-to-Real Transfer Kim, Donghyeon,Berseth, Glen,Schwartz, Mathew,Park, Jaeheung* Conference: 2024 IEEE International Conference on Robotics and Automation (ICRA). In this new work we find the deepRL and torque control make a very good combination! Torque control is normally under appreciated because it is more intuitive to make position controlled planning systems. However, we find that with a unique type of torque pretraining for gravity compensation, we are able to train torque-based controllers faster and gain the benefits of compliance which is a large bonus for Sim2Real trasnfer. Video Link: https://lnkd.in/g4gecaUD Paper link: https://lnkd.in/gfSTMv8Q

50

📣 Kiroong Choe is presenting our new IEEE TVCG paper, Enhancing Data Literacy On-demand: LLMs as Guides for Novices in Chart Interpretation, at the "Look, Learn, Language Models" session this Friday: https://lnkd.in/eygdA_jt 📝 Author List: Kiroong Choe, Chaerin Lee, Soohyun Lee, Jiwon Song, Aeri Cho, Nam Wook Kim, Jinwook Seo 🎓 Kiroong (https://kiroong.github.io/), a Ph.D. student at Seoul National University advised by Prof. Jinwook Seo, spearheaded this work by leading a group of other students. I enjoy meeting with him because he always thinks critically about my input and offers different perspectives when needed, which in turn improves my own thinking. 🙂 📊 We compared three different methods of learning assistance for interpreting complex visualizations for novices: 1) using web search, 2) LLMs with text-based support, and 3) LLMs with both text and visual support. We developed a "visual prompting" system to support the third condition. 💡 An interesting finding is that, despite participants preferring the LLM-based approach, when we counted the number of insights, the web search performed the best, even though it was more time-consuming and challenging to use. 🤔 This might suggest that cognitive difficulty leads to more engagement and thus more learning. This would be an important consideration for future LLM-based learning systems. 📄 Please check out the full paper manuscript: https://lnkd.in/etfn6yGg

17

It was an honor to host Jian-Xun Wang from University of Notre Dame at our data-driven physical simulation (#DDPS) seminar, Lawrence Livermore National Laboratory, on September 19, 2024. He gave a wonderful talk on "#Neural #Differentiable #Physics: Unifying Numerical PDEs and Deep Learning for Data-Augmented Computational Physics." You can watch his talk again through https://lnkd.in/eAcAzCy7 #Abstract: Predictive modeling and simulation are crucial for understanding and controlling complex physical processes across engineering disciplines. Traditional numerical models, based on first principles, struggle with complex systems involving multiple physics across varied scales due to incomplete understanding and inadequate mathematical models. Additionally, traditional solvers' high computational demands hinder real-time control and repeated model queries in design optimization and uncertainty quantification. However, advances in sensing technology, data availability, machine learning, and GPU computing offer new opportunities for data-driven models that improve predictive capabilities. Despite these advances, challenges like data requirements, limited generalizability, and interpretability persist. This talk introduces our efforts in Neural Differentiable Physics, a SciML framework unifying numerical PDE solvers and deep learning for modeling complex physical systems. By integrating known physics, multi-resolution data, numerical techniques, and neural networks, this novel approach offers a departure from existing frameworks like PINNs. Combining physical principles with cutting-edge AI promises significant advancements in understanding and modeling complex systems with broad applications in science and engineering. Useful links: 📚 DDPS seminar is organized by libROM team ( www.librom.net ). 🗓️ DDPS seminar schedule: https://lnkd.in/gnGsx6qq 💻 LLNL News: https://www.llnl.gov/news

149

4 Comments

The paper from my MIDS program project at UC Berkeley School of Information is now published. Thanks to everyone who worked on this project. You can access it here: https://lnkd.in/g4pzfvwU #ai #deeplearning #machinelearning #materialsscience #berkeley #ischool #metamaterials

62

7 Comments

Numerical Integration with Simpson's 3/8 Rule in MATLAB In this tutorial, I'll show you how to use Simpson's 3/8 Rule to numerically integrate functions in MATLAB. This rule is a more accurate version of the Simpson's 1/3 Rule, and it's particularly useful for integrating functions that have a more complex shape. Key topics covered in the video: What is Simpson's 3/8 Rule? How to implement Simpson's 3/8 Rule in MATLAB Examples of using Simpson's 3/8 Rule to solve real-world problems Watch the video to learn more: I hope you find this video helpful! If you have any questions, please leave a comment below. I would also appreciate it if you could share this video with your friends and colleagues. Thank you for watching! #NumericalIntegration #MATLAB #Simpsons38Rule #Engineering #Mathematics #Tutorial #Education #OnlineLearning #StudyTips #MathProblems

13

Happy to share that our paper "On Optimal Server Allocation for Moldable Jobs with Concave Speed-Up" has been accepted for presentation at ACM Mobihoc 2024, a highly selective single-track conference on Theory, Algorithmic Foundations, and Protocol Design for Mobile Networks and Mobile Computing. It addresses the fundamental question of how to optimally allocate computing resources to parallel tasks including machine learning workloads running on clouds. It uses Stein's method to derive asymptotically optimal algorithms for resource allocation for parallel tasks. The arXiv version can be found in the link given below. https://lnkd.in/dwZeMhzk

19

2 Comments

A substantial piece of work from Apple led by Iman Mirzadeh, keivan alizadeh vahid, Oncel Tuzel, Samy Bengio, Mehrdad Farajtabar and Hooman Shahrokhi supports the thesis that transformer-based LLMs do not reason--they perform pattern-matching. The team from  also provides a great benchmarking tool--GSM-Symbolic for assessing mathematical and formal reasoning. This supports my anecdotal findings when I evaluated o1 https://lnkd.in/ec72fecb as well as my thoughts on the incompleteness of our assessments of AGI https://lnkd.in/ehBT3HB3

16

2 Comments

Our paper "EffFix: Efficient and Effective Repair of Pointer Manipulating Programs " has been recently accepted for publication in ACM Transactions on Software Engineering and Methodology (TOSEM). Kudos to Yuntong Zhang, Andreea Costea, Davin McCall, Abhik Roychoudhury. This work proposes a novel program repair technique using probabilistic context-free grammar guided by static analysis. The key idea is to integrate static analysis and program repair such that the repair engine can incrementally learns what a correct patch may look like based on its effect on the symbolic heap. Experimental evaluation of fixing real world memory errors in medium to large scale subjects like OpenSSL and Linux Kernel, shows the efficiency and effectiveness of EffFix. In particular, EffFix has a fix ratio of 66% for memory leak bugs and 83% for Null Pointer Dereferences for the considered dataset. Pre-print: https://lnkd.in/gRrV43Pc #publication #TOSEM #programrepair

46

Call for Papers: TPAMI Special Issue on "Generative AI in 3D Vision" Dear Colleagues, We are pleased to announce a Special Issue on "Generative AI in 3D Vision" in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI). This special issue aims to bring together cutting-edge research in the field of 3D generative models and their applications. Topics of interest include but are not limited to: Advances in 3D generative models, such as GANs, VAEs, Autoregressive models, Diffusion models, Normalizing Flow, etc. Innovations in 3D representations for GenAI, such as point clouds, meshes, voxels, neural fields, Gaussian splatting, multiplane images, etc. 3D reconstruction from single or sparse views Novel view synthesis 3D human generation, reconstruction, and animation 3D from multi-modality, such as text, audio, etc. Multi-view image generation from multi-modality, such as text, audio, etc. 4D (dynamic) generation and reconstruction 3D scene/object manipulation, such as instruction-based editing, drag-based editing, multi-object interaction, style transfer, etc. Applications of 3D generative models in fields like virtual/augmented reality, gaming, film production, computer-aided design, Metaverse, robotics, and embodied AI Large-scale 3D datasets Evaluation metrics and benchmarks for 3D generative models Efficiency and scalability of 3D generative models Robustness and security of 3D generative models Important Dates: Manuscript submission deadline: Oct 15, 2024 First review notification: January 15, 2025 Revised manuscript submission: March 15, 2025 Final review notification: May 15, 2025 Final manuscript submission: June 15, 2025 Publication date: August 15, 2025 For more information, please visit https://genai3d.github.io. We look forward to your submissions! Best regards, Deqing Sun, Xiangyu Xu, Siyu Tang, Angela Dai, Jiajun Wu, Katerina Fragkiadaki, Andrea Tagliasacchi, Christian Theobalt, and Ming-Hsuan Yang

146

4 Comments

Excited to share that my research article titled "Enhancing the Performance of Extreme Learning Machine Technique Using Optimization Algorithms for Embedded Workload Characterization" has been published in the Alexandria Engineering Journal (Volume 108, 2024, Pages 509-517). This work explores the use of hybrid optimization algorithms like PSO, GA, ACO, and WOA to enhance the accuracy of ELM for embedded systems, achieving 98.5% classification accuracy. Special thanks to my co-authors, especially SARAVANA KUMAR R and collaborators. Check out the full paper here: https://lnkd.in/dQ5BYAJ9 #Research #EmbeddedSystems #MachineLearning #Optimization #IoT #AI #AcademicPublication

33

3 Comments