Stefano Ermon

I am pleased to present my advanced microeconomics lecture, in which I provide a detailed, step-by-step guide on solving Utility Maximization problems using the Cobb-Douglas Utility Function. This comprehensive tutorial is designed to enhance your understanding of this essential concept in microeconomic theory. Throughout the lecture, I break down the problem-solving process into clear, manageable steps, ensuring that you grasp the fundamental principles and techniques required to tackle these complex problems with confidence. By the end of this lecture, you will have gained valuable insights into the practical application of the Cobb-Douglas Utility Function and its significance in consumer choice theory. Whether you are a student seeking to deepen your knowledge of advanced microeconomics or a professional looking to refresh your understanding of this crucial topic, this lecture is an invaluable resource. I invite you to engage with the material, ask questions, and explore the fascinating world of utility maximization. Embark on this educational journey with me and unlock the power of the Cobb-Douglas Utility Function in solving Utility Maximization problems. Your understanding of microeconomic theory will undoubtedly reach new heights.

3

Quantum Internet - Please check our paper published on IEEE Network magazine. https://lnkd.in/e7ueWYVW Uman Khalid Junaid ur Rehman, Ph.D. Saw Nang Paing Haejoon Jung We discuss the potential of the near-term quantum networks using noisy intermediate-scale quantum (NISQ) technologies. It highlights the architectural and operational principles necessary for developing these quantum networks and emphasizes their application in enhancing quantum computing, communication, and sensing. The paper provides a comparative analysis of network performance improvements and addresses technical challenges that must be overcome to realize practical and effective quantum networks. Overall, the paper identifies the advancements and ongoing efforts towards achieving a functional quantum internet, delineating a clear path from current capabilities to future possibilities.

69

1 Comment

Our latest paper, "Distributed Matching-by-Clone Hungarian-Based Algorithm (DMCHBA) for Task Allocation of Multi-agent Systems," has just been published in IEEE Transactions on Robotics! This work introduces a scalable distributed task allocation algorithm designed to operate efficiently across multi-agent systems with any communication network topology. This paper has been selected for presentation at the 2024 IEEE International Conference on Robotics and Automation in Japan! DMCHBA is tailored for complex multi-agent environments, ensuring flexibility across various network topologies. Through rigorous Monte Carlo simulations, DMCHBA has demonstrated superior performance in terms of both overall cost and runtime compared to state-of-the-art algorithms. We provide a detailed analysis of the algorithmic computational complexity and communication complexity, reinforcing the robustness of DMCHBA. This achievement is a testament to the hard work and dedication of Arezoo Samiei, PhD. We believe that these findings will make a significant impact on task allocation processes in robotic systems, pushing the boundaries of what's possible in automated environments. Read the paper here: https://lnkd.in/g7RAp5fS. We are working to publish the code on MATLAB File Exchange. More to come!

29

6 Comments

I am thrilled to announce that one of my major PhD contributions with Prof. Kamil Khan, "Computing subgradients of convex relaxations for solutions of parametric ordinary differential equations", has been published in Optimization Methods and Software! https://lnkd.in/e625D2H4 For aiding in global dynamic optimizaiton, we proposed new sensitivitiy evaluation approaches for state-of-the-art nonsmooth convex relaxations of dynamic systems. Moreover, we for the first time extended the classical adjoint sensitivity evaluation approach for smooth dynamic systems to computing sensitivities for nonsmooth convex dynamic systems. Due to the success of adjoint sensitivity approaches in local dynamic optimization, we believe this result can significantly accelerate lower bounding in global dynamic optimization. Indeed, the mentioned new adjoint approach has been implemented in C++, see here: https://lnkd.in/e-YqFh4s

46

ALHAMDULILLAH! Excited to share that our paper on "Cervical Spine Fracture Detection and Classification Using a Two-Stage Deep Learning Methodology" has been published in IEEE Access! This research represents a significant step in improving the accuracy and efficiency of diagnosing cervical spine fractures, which is crucial for timely and effective treatment. I am grateful to all the collaborators and supporters who have contributed to this work. Looking forward to the impact this research will have on AI in the medical field. #CervicalSpine #DeepLearning #MachineLearning #ComputerVision #MedicalResearch #PublishedPaper

5

1 Comment

I am pleased to share that my paper titled "Exploring the Different Order of Switches During Multi-switching Synchronization" has been published in Iranian Journal of Science. https://lnkd.in/gJ2V6eN5

13

5 Comments

I'm happy to share that our paper titled “Gegenbauer Graph Neural Networks for Time-Varying Signal Reconstruction” has been accepted for publication in IEEE Transactions on Neural Networks and Learning Systems. You can access the paper here: https://lnkd.in/eGNfPc_a and ArXiv link: https://lnkd.in/e-mqmqgp. This work is part of our recent efforts in the broader field of spatiotemporal graph analysis. In our study, we introduce two significant contributions. Firstly, we propose a specialized Graph Neural Network (GNN) architecture designed specifically for addressing the task of time-varying graph signals recovery. Secondly, we present a novel graph convolution approach based on Gegenbauer polynomials, which generalizes the well-known Chebyshev convolutions in GNNs. Our research marks a notable advancement in the domain of time-varying graph signals, a field predominantly explored through the lens of graph signal processing. Through experimentation, we discovered that utilizing traditional methods like the Graph Convolutional Network (GCN) for this task is inadequate. Instead, we developed an autoencoder-type architecture that achieves competitive performance. This paper is the result of a fruitful collaboration with Jhon Castro-Correa (first author), Mohsen Badiey, and Fragkiskos Malliaros. If you're interested in delving deeper into this topic, feel free to explore our GitHub repository https://lnkd.in/evmRdRxU. Special thanks to all the institutions involved in supporting this research Télécom Paris, Institut Polytechnique de Paris, University of Delaware, CentraleSupélec, Université Paris-Saclay, Inria. I welcome any discussions and look forward to sharing more insights in the future. Stay tuned for further developments in our exploration of spatiotemporal graphs! #gnn, #graph, #spatiotemporalgraphs, #gnnupdates

61

6 Comments

If you're interested in GNNs and spatiotemporal graph prediction tasks, check out our recent IEEE TNNLS paper!

26

#Paper_accepted Excited to share that our latest research paper titled 'Interactive singing melody extraction based on active adaptation' has been accepted for publication in the IEEE/ACM Transactions on Audio, Speech, and Language Processing (TASLP)! The preprint is available here - https://lnkd.in/g4rA7jgm   The existing works in melody extraction focus on training a model on data from the source domain, and testing/deploying it on the target domain. The performance of such models can be boosted by adapting the model using very little annotated data from the target domain. This work proposes a novel model-agnostic active-meta-learning domain adaptation algorithm that combines active learning and meta-learning. Given a spectrogram of an audio in the target domain, the model uses active learning to select those frames of the spectrogram where it is least confident and marks those frames for the human annotator to annotate. Once the human annotator provides the melody annotations for those frames, the model uses meta-learning to adapt its parameters. In this way, the model adapts to the target domain and improves the performance.   A big thank you to the reviewers and the editorial team for their valuable feedback and support.

8

Happy to share a publication update from our group MISN IITD.  Our work “Optimization Framework for Semi-supervised Attributed Graph Coarsening” got accepted to UAI2024.One of the first approach for Optimization based semi-supervised graph coarsening. Abstract: In data-intensive applications, graphs serve as foundational structures across various domains. However, the increasing size of datasets poses significant challenges to performing downstream tasks. To address this problem, techniques such as graph coarsening, condensation, and summarization have been developed to create a coarsened graph while preserving important properties of the original graph by considering both the graph matrix and the feature or attribute matrix of the original graph as inputs. However, existing graph coarsening tech- niques often neglect the label information during the coarsening process, which can result in a lower quality coarsened graph and limit its suitability for downstream tasks. To overcome this limitation, we introduce the Label-Aware Graph Coarsening (LAGC) algorithm, a semi-supervised approach that incorporates the graph matrix, feature matrix, and some of the node label information to learn a coarsened graph. Our proposed formulation is a non-convex optimization problem that is efficiently solved using block successive upper bound mini- mization(BSUM) technique, and it is provably con- vergent. Our extensive results demonstrate that the LAGC algorithm outperforms the existing state-of- the-art method by a significant margin. Congratulations! Manoj kumar Subhanu Halder#UAI #graphML

119

6 Comments

My Machine Learning course in Bengali language will be running once again on NPTEL, during June-September 2024! Please join in large numbers, and also spread the word in your circles! https://lnkd.in/drm8UtuJ

137

2 Comments

Recently, I gave a keynote talk on the optimal deployment of RIS in mmWave cellular networks at EECS Research Students Symposium at IISc that showcases work done by PhD students. Full paper is available at https://lnkd.in/gF_6W_yp. In this paper, we discuss the design of RIS locations in a cellular network with blockages.

141

Our IEEE Computer Society Computing in Science & Engineering (CiSE) article on Enhancing Scientific Research with FAIR Digital Objects in the National Science Data Fabric has been published. It is available for open access to all. Download the article here: https://lnkd.in/g_4CrGgu Learn more about our vision of combining findable, accessible, interoperable, and reusable (FAIR) Digital Objects with the National Science Data Fabric (NSDF) to enhance data accessibility, scientific discovery, and education. Integrating FAIR Digital Objects into the NSDF overcomes data access barriers and facilitates the extraction of machine-actionable metadata in alignment with FAIR principles. The article discusses examples of climate simulations and materials science workflows. It establishes the groundwork for a dataflow design that prioritizes inclusivity, web-centricity, and a network-first approach to democratize data access and create opportunities for research and collaboration in the scientific community. Contact us if you want to collaborate or share your thoughts about data democratization. Michela Taufer Valerio Pascucci Christine Kirkpatrick Global Computing Laboratory San Diego Supercomputer Center Scientific Computing and Imaging Institute at the University of Utah University of Tennessee, Knoxville National Science Foundation (NSF)

47

I am happy to share my latest paper published in IEEE Transactions on Instrumentation and Measurement: https://lnkd.in/exjwx5j9 Thanks to Dr Gourav Bhattacharya, Ms Indrianita Lionadi and Dr Joanna Ward as the coauthors of this work. In this work, we investigated and compared the impact of laser positioning on the captured dynamic responses of both standard and our proposed modified microcantilevers which we published recently in Advanced Science. Subsequently, we assessed the imaging performance of different eigenmodes of tailored modified cantilever by locating laser spots at different positions of the cantilever surface. Our study offers guidelines for optimizing the spot position to enhance quantification and imaging using tailored cantilever eigenfrequencies.

58

2 Comments

I am happy to announce that our paper, "Probabilistic Physics-Integrated Neural Differentiable Modeling for Isothermal Chemical Vapor Infiltration Process," has been officially published in npj Computational Materials. This research was led by my grad student Deepak Akhare, in collaboration with Zeping Chen, Prof. Tengfei Luo and Dr. Richard Gulotty from #Honeywell. Our work focuses on modeling the manufacturing process of carbon-carbon composites, which significantly influences their performance, quality, and consistency. Given the physical complexities and limited experimental data of the CVI densification process, we developed a data-driven predictive model that seamlessly integrates backbone governing PDEs and trainable neural operators through differentiable programming. An ensemble-based uncertainty quantification feature enhances the model’s reliability and robustness. Through comprehensive numerical experiments with both synthetic and real-world data from #Honeywell, our method demonstrates its capability in predicting densification during the CVI process. This research highlights another successful application of our neural differentiable modeling framework, which unifies physics-based numerical PDEs and deep learning via #differentiableprogramming, allowing for effective data-driven modeling of complex system by integrating partially known physics and sparse/indirect measurements. More details can be found https://lnkd.in/gCwxAQSh #CVI #CoMSAIL #NeuralDifferentiableModeling #SciML #AI4Science

88

Perturbation theory for your arbitrary wavefunctions: https://lnkd.in/egaKPQw3 Brought to you by LOKHANDE RUGWED and Pratiksha Gaikwad #computationalchemistry #perturbationtheory #coupledcluster

40

1 Comment

Research Associate Anand Prakash Tiwari, PhD student Saifur Rahman, and Professor Will Scheideler co-authored a paper that presents a novel 3D printing method to create low-cost and efficient electrodes for electrocatalytic hydrogen production. The paper, entitled "3D Printed Microlattices of Transition Metal/Metal Oxides for Highly Stable and Efficient Water Splitting," was published in Advanced Materials Technologies. "This method, called polymer infusion additive manufacturing (PIAM), allows for the creation of large-area, 3D-printed structures with carbon core and metal/metal oxide shells. These structures outperform state-of-the-art metallic foams and can be easily customized with various transition metals, such as copper or cobalt, to optimize their performance. The resulting materials have shown exceptional durability and electrocatalytic activity, making them promising for large-scale water splitting and sustainable hydrogen fuel production."—Saifur Rahman Read more: https://bit.ly/3UD3qdf

37

Glad to share our group's recent research from the Indian Institute of Science (IISc) on fast hardware-aware matrix-free algorithms for finite-element discretized sparse matrix multivector products on multi-node CPU and GPU architectures published in the "Journal of Parallel and Distributed Computing (JPDC)". We address a critical gap in existing matrix-free algorithms that are not well suited for the action of a finite-element discretised sparse matrix on large number of vectors. We rely on carefully reformulating this global sparse matrix-multivector problem into a sequence of smaller-sized dense tensor contractions and devising architecture-specific batched evaluation strategies to overlap compute with data movement for achieving efficient pipelining and reduced data access in a distributed setting. Using the proposed approach, we observe substantial computational gains compared to the state-of-the-art algorithms and libraries when dealing with large number of vectors and demonstrate the usefulness in accelerating large-scale eigenvalue problems. These methods and algorithms have implications in accelerating iterative solvers involving the solution of large-scale sparse eigenvalue problems, linear systems of equations with multiple right-hand sides, and also possibly accelerating algorithms involving sequence of small dense tensor contractions (eg: ML backpropagation) on emerging computing architectures. Read more about our work at https://lnkd.in/gMRiJtHD Hearty congratulations to my PhD students Gourab Panigrahi and Nikhil Kodali, who led the work, and Debashis Panda, a former project associate at IISc.

166

2 Comments

Here you will find my two recently published journals, *Unified predictive modeling of supercontinuum spectra: Using multi-material data with Closed-Form Continuous Time Neural Networks-Published in 'Optics Communication(Elsevier)' https://lnkd.in/dMrJZAB8 *Employing deep learning for predicting supercontinuum generation in chalcogenide planar waveguide-Published in 'Optik(Elsevier)'. https://lnkd.in/dMtF6eAw

20

3 Comments