International Union of Crystallography’s Post

📅 Workshop on Scientific Machine Learning – October 2-4, 2024 📅 The "𝐖𝐨𝐫𝐤𝐬𝐡𝐨𝐩 𝐨𝐧 𝐒𝐜𝐢𝐞𝐧𝐭𝐢𝐟𝐢𝐜 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠" is hosted by the Oden Institute for Computational Engineering and Sciences at The University of Texas at Austin. This second annual workshop, led by Stella Offner and Tan Bui-Thanh, will explore the role of 𝐒𝐜𝐢𝐞𝐧𝐭𝐢𝐟𝐢𝐜 𝐌𝐚𝐜𝐡𝐢𝐧𝐞 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 (𝐒𝐜𝐢𝐌𝐋) in solving complex problems across science, engineering, and medicine. This workshop will bring together leading experts in computational science and machine learning to share cutting-edge research, identify key challenges, and establish new directions in 𝐒𝐜𝐢𝐌𝐋. 🔗 Add to Your Schedule: https://lnkd.in/gjTNxy8Q #simulAItion #Simulation #AI #Innovation #SciML #ComputationalScience

🚀 Excited to share insights from my recent research on the landscape of local optima in Maximum Satisfiability (Max-SAT) problems! 🧠 In this study, we explored how the configuration of local optima and their neighbors influences solution strategies in complex optimization scenarios, specifically focusing on Max-SAT. Our findings reveal that as the complexity of the problem increases—by adding more variables per clause or more clauses—the local search space becomes denser with optima that are nearly as good as the best solutions. This indicates a more challenging environment for distinguishing and selecting the optimal solution. 🔍 Key Takeaways: Increased Problem Complexity: Leads to a higher percentage of neighboring solutions that match the height of their corresponding local optima, suggesting denser clusters of effective solutions. Impact on Algorithms: These results emphasize the need for more sophisticated search strategies in optimization algorithms used within machine learning frameworks, particularly in areas such as feature selection, hyperparameter tuning, and model optimization. 🤖 Implications for Machine Learning: Understanding these optimization landscapes is crucial for developing more efficient machine learning algorithms. By enhancing our algorithms' ability to navigate through dense clusters of local optima, we can significantly improve their performance, especially in learning scenarios that involve large and complex datasets with numerous features. This research not only advances our knowledge in theoretical computer science but also has practical applications in machine learning, providing a foundation for more robust and efficient algorithmic strategies. Excited to see how this can push the boundaries of what's possible in AI and machine learning! #MachineLearning #ArtificialIntelligence #Optimization #Research #DataScience

📅 AI Accelerating Scientific Understanding – May 10, 2024 📅 Vector Institute Distinguished Talk series brings together academic and industrial data scientists to discuss advanced machine learning topics. In the upcoming talk, "𝐀𝐈 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐢𝐧𝐠 𝐒𝐜𝐢𝐞𝐧𝐭𝐢𝐟𝐢𝐜 𝐔𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠: 𝐍𝐞𝐮𝐫𝐚𝐥 𝐎𝐩𝐞𝐫𝐚𝐭𝐨𝐫𝐬 𝐟𝐨𝐫 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐨𝐧 𝐅𝐮𝐧𝐜𝐭𝐢𝐨𝐧 𝐒𝐩𝐚𝐜𝐞𝐬," 𝘗𝘳𝘰𝘧. 𝘈𝘯𝘪𝘮𝘢 𝘈𝘯𝘢𝘯𝘥𝘬𝘶𝘮𝘢𝘳 will explore AI-based simulation methods that are significantly faster and cheaper than traditional simulations. These methods, based on 𝐍𝐞𝐮𝐫𝐚𝐥 𝐎𝐩𝐞𝐫𝐚𝐭𝐨𝐫𝐬, have revolutionized fields such as weather forecasting, fluid dynamics, and medical device design. The event will be held virtually on Friday, May 10, 2024, from 12:30 to 13:30 (GMT-4), and it is open to the public. 🔗 Add to Your Schedule: https://lnkd.in/gViftg9x #simulAItion #Simulation #AI #Innovation #NeuralOperators

Pretty exciting announcement from DeepMind’s latest AI systems, AlphaProof and AlphaGeometry 2, have achieved remarkable success by solving four out of six problems from the International Mathematical Olympiad (IMO), reaching a silver medal level! 🥈 𝗞𝗲𝘆 𝗧𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀: 𝟭. 𝗔𝗹𝗽𝗵𝗮𝗣𝗿𝗼𝗼𝗳 𝗮𝗻𝗱 𝗔𝗹𝗽𝗵𝗮𝗚𝗲𝗼𝗺𝗲𝘁𝗿𝘆 𝟮: These AI systems tackled complex IMO problems, demonstrating advanced mathematical reasoning. 𝟮. 𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲: Scoring 28 out of 42 points, these models performed at the top end of the silver-medal category. 𝟯. 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗮𝗻𝗱 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁: AlphaProof was trained by proving millions of problems, while AlphaGeometry 2 was enhanced with more synthetic data. 𝟰. 𝗙𝘂𝘁𝘂𝗿𝗲 𝗣𝗼𝘁𝗲𝗻𝘁𝗶𝗮𝗹: These tools aim to assist mathematicians in exploring hypotheses, solving long-standing problems, and completing proofs quickly. Let’s discuss and explore how these models can transform our industries. 💬 https://lnkd.in/eCP3phPh

My first research work by myself : Beyond SVM - Designing a Scalable and Robust Model for Classification and Regression Tasks 🚀 I am excited to share my latest research work on advancing Support Vector Machines (SVM) and addressing some of its key limitations. This paper proposes a novel approach to improve scalability, robustness, and interpretability in machine learning models. 📄 Abstract: In this paper, I explored the inherent challenges of Support Vector Machines, including their computational cost, sensitivity to noise, and lack of interpretability. We propose enhancements that incorporate an adaptive margin, dynamic kernel optimization, and stochastic training to overcome these limitations, making the model more scalable and effective for large datasets. 🌟 Key Contributions: Adaptive Margin: Introducing a dynamically adjustable margin to improve the model’s flexibility in handling varying data densities. Kernel Optimization: Implementing an adaptive kernel function that evolves during training to better fit data distributions. Stochastic Gradient Descent: Replacing the computationally expensive quadratic programming (QP) solver to handle large datasets efficiently. Improved Interpretability: Enhancing model transparency for better understanding and decision-making. Significance: This research paves the way for next-generation classification and regression models, which can be applied in diverse fields such as finance, healthcare, and e-commerce. By combining traditional machine learning approaches with modern techniques, we aim to deliver a model that is not only more powerful but also scalable and easier to interpret. You can read the full paper here I’d love to hear your thoughts, feedback!! #MachineLearning #AI #ResearchWork

APDIO’s Tutorial, The Interplay between Machine Learning and Mathematical Optimization, Cristina Molero Following the workshop “Bridging the Gap: APDIO's Workshop on OR and AI” on Monday, September 16, APDIO will host a tutorial by Cristina Molero on Tuesday, September 17, titled "The Interplay between Machine Learning and Mathematical Optimization”.   In this tutorial, the use of mathematical optimization to support the construction of machine learning models will be explored, including: - Classification and regression trees - Ensemble methods - Rule sets - Risk scores - Clustering methods   If time permits, we will also discuss how machine learning can be used to solve mixed-integer nonlinear optimization problems.   Target audience: master and PhD students in optimization, artificial intelligence and related areas   Requirements: basic knowledge on optimization and machine learning    Registration until September 6. A confirmation e-mail will be sent on September 9, due to limited availability for the Tutorial: https://lnkd.in/dVaQS9tk

Have you ever wondered if there’s a roadmap to the universe’s greatest puzzles? Picture this: a single map that unveils the secrets of the cosmos, unlocking the mysteries of numbers, shapes, and patterns that govern our world. In honor of mathematics awareness month, let me introduce the "Map of Mathematics" – a mesmerizing infographic that serves as the foundation for the era of AI, making it a starting point for those aspiring to venture into artificial intelligence. From the enchanting arithmetic and algebra to calculus and beyond, this map guides you through the landscape of mathematical thought, laying the groundwork for understanding the algorithms and principles that drive AI innovation. So, to those who often ask me where to start their journey into AI, I say: get familiar with math first. It's the bedrock upon which the wonders of AI are built. #MathAwarenessMonth #AI #DataScience #Stem Subscribe to #CXOSpice newsletter (https://lnkd.in/gy2RJ9xg) and tune in for more insights about how growth thrives at the intersection of Technology and Humanity. Subscribe to #CXOSpice Youtube channel (https://lnkd.in/gnMc-Vpj), check out the latest episode on "Spark Innovation with Copilot M365" featuring Microsoft and tune in for the upcoming episode with Jen Larson at Intel. Follow me on X @YuHelenYu (https://lnkd.in/gD9Ti6fJ) to gain additional insights on #Technology and #Innovation. Image Credit: Domain of Science

🚀 Calling All Machine Learning on Graphs Enthusiasts! I am thrilled to share that I, along with my fellow organizers—FAROOQ AHMAD WANI , Maria Sofia Bucarelli, and Giulia Di Teodoro—am organizing the #IJCNN2025 Competition: Learning with Noisy Graph Labels! 🎉 Submissions are now open, and we invite researchers, practitioners, and students to tackle one of the biggest challenges in graph learning: handling noisy labels 🎯. Why Participate? ✅ Real-World Impact: Propose effective methods that can benefit existing domains such as biology, finance or social networks where graph data may be mislabelled. ✅ Advance Research: Contribute to a relatively unexplored yet critical area in Graph Neural Networks. ✅ Collaborate with a Community: Join us in pushing the boundaries of machine learning. 🏆 Reward: Top teams can submit a paper for consideration in the IJCNN 2025 Proceedings on IEEE Xplore. Timeline 📅 Submission Opens: December 23, 2024 📅 Submission Deadline: February 10, 2025 📅 Winners Announced: February 15, 2025 📅 Winners' Paper Submission Deadline: March 1, 2025 🌐 Visit our competition homepage for more details: https://lnkd.in/dXtuMzca We, the organizing team, can’t wait to see your creative approaches and groundbreaking ideas. Let’s push the boundaries of graph learning under label noise together! 💡✨ #IJCNN2025 #MachineLearning #GraphNeuralNetworks #AICompetition #NoisyLabels

🚀 Exciting News! 🚀 I’m happy to share that our paper, "GSP-KalmanNet: Tracking Graph Signals via Neural-Aided Kalman Filtering," has been accepted to the IEEE Transactions on Signal Processing! 🏆 In this work, we tackled the challenge of tracking dynamic systems of graph signals in complex domains like transportation networks and power grids. ⚡️ By combining Graph Signal Processing and Deep Learning techniques, we developed GSP-KalmanNet, which enhances accuracy, runtime performance, and robustness compared to traditional methods. 🌟 Check out the full paper here: https://lnkd.in/dxxUY5xb Proud to have achieved 3️⃣ accepted journal papers during my Master's journey! 🎓📈 #Research #DeepLearning #GraphSignalProcessing

I am happy to announce the publication of my latest article on the use of machine learning to predict country-level CO2 emissions. The study "Determinants of CO2 Emissions: A Machine Learning Approach" published in International Advances in Economic Research presents research that directly compares the predictive capabilities of machine learning models with traditional econometric models. Key Insights: 1️⃣ Advanced Machine Learning: Artificial Neural Networks (ANN) outperformed all other models, showcasing incredible precision with a 98% R-squared value on the test dataset. 2️⃣ Critical Variables Identified: GDP per capita and agriculture’s value-added contribution are significant determinants of per capita CO2 emissions. 3️⃣ Enhanced Predictive Power: Machine learning, particularly ANN, holds immense potential for improving our forecasting abilities, offering a robust tool for policymakers. ESSCA Research Marie Le Borgne-Larivière (Costes) Silke Leukefeld IPCC Valuation and Sustainability #AIforSustainability #EnvironmentalPolicy #MachineLearning #ESSCA