evoxels: Differentiable Physics for Voxel-Based Microstructure Simulation Published

𝐒𝐚𝐦𝐞 𝐞𝐪𝐮𝐚𝐭𝐢𝐨𝐧. 𝐒𝐚𝐦𝐞 𝐢𝐧𝐢𝐭𝐢𝐚𝐥 𝐜𝐨𝐧𝐝𝐢𝐭𝐢𝐨𝐧. 𝐂𝐨𝐦𝐩𝐥𝐞𝐭𝐞𝐥𝐲 𝐝𝐢𝐟𝐟𝐞𝐫𝐞𝐧𝐭 𝐩𝐡𝐲𝐬𝐢𝐜𝐬. As part of our High-Speed CFD Code Series, I explored Burgers’ equation in two ways: 1️⃣ Non-conservative form → u uₓ 2️⃣ Conservative form → ∂ₓ(u²/2) Mathematically identical, right? Numerically… not even close. 🚨 What happens: • Non-conservative → mass is NOT preserved • Solution gets artificially diffused • Shock behavior becomes unreliable ✅ Conservative (flux form): • Exact discrete conservation • Physically correct transport • Proper shock behavior This is the moment every CFD learner realizes: 👉 “The PDE is not the problem. The discretization is.” In this tutorial, I’ve included: • Cole–Hopf analytical solution • 1D and 2D simulations • Mass conservation diagnostics • Error analysis 🔗 Full tutorial: https://lnkd.in/gGT5s_ec The tutorial series will cover problems ranging from simple convection equations to 2-D supersonic flow over a wedge, all implemented in Python. We will see how to set up a differentiable solver and optimize it both with and without neural networks at the end. Please explore the tutorial and share feedback and mistakes. If you're working on CFD, numerical PDEs, or scientific computing in Python, I’d love to hear your thoughts or suggestions! 𝘚𝘩𝘢𝘭𝘭 𝘸𝘦 𝘴𝘵𝘢𝘳𝘵 𝘭𝘦𝘢𝘳𝘯𝘪𝘯𝘨 𝘰𝘱𝘵𝘪𝘮𝘪𝘻𝘢𝘵𝘪𝘰𝘯 𝘢𝘯𝘥 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘵𝘪𝘢𝘭 𝘱𝘳𝘰𝘨𝘳𝘢𝘮𝘮𝘪𝘯𝘨 𝘧𝘳𝘰𝘮 𝘩𝘦𝘳𝘦, 𝘰𝘳 𝘴𝘩𝘢𝘭𝘭 𝘐 𝘤𝘰𝘷𝘦𝘳 𝘪𝘵 𝘢𝘵 𝘵𝘩𝘦 𝘦𝘯𝘥? 𝘗𝘭𝘦𝘢𝘴𝘦 𝘴𝘩𝘢𝘳𝘦 𝘺𝘰𝘶𝘳 𝘧𝘦𝘦𝘥𝘣𝘢𝘤𝘬. #CFD #NumericalMethods #FluidDynamics #ScientificComputing #PhysicsInformedAI

🚀🔬 Exciting news for tech leaders and data enthusiasts! Dive into the world of computational materials science with our latest find – a tutorial using the powerful pymatgen library in Python. This tool offers a unique opportunity to build and analyze crystal structures, such as silicon, sodium chloride, and even complex LiFePO₄-like materials! 🔍 But that's not all! With this implementation, you can investigate lattice properties, densities, compositions, and even symmetry analysis using space-group detection. You'll also be able to examine atomic coordination environments and apply oxidation states. 💪 Why is this important? As we continue to push the boundaries of technology, understanding materials at a molecular level will play a crucial role in designing more efficient electronics, batteries, and materials for various industries. Embrace the future and stay ahead of the curve with pymatgen! 💡 Don't miss out on this opportunity to expand your knowledge and skills in AI for materials science. Share your thoughts on how you see this technology impacting the industry and let's shape the future together. #AI #MaterialScience #DataAnalytics #Python #TechLeadership #Innovation

Huge congrats to Houston Haynes on his latest publication! Real expertise means tuning for performance through deep inspection, questioning, and evaluation. Over the past few months (I've loved those text updates!), Houston developed a formal framework for a problem most of us have not encountered: how to preserve the semantic meaning of code, including dimensional context like units of measure, through every stage of compilation instead of discarding it early. This lets the compiler make more precise decisions about numeric representation, memory placement, and cross-target performance trade-offs. It also surfaces those decisions to engineers as real-time design feedback. It's a thoughtful look at where high-performance computing and programming languages are heading, and how much opportunity remains in translating intent into execution. Read more below

Twelve years ago, this work lived in my research codebase in MATLAB. Today, it's back as a working interactive web demo. This demo brings to life the method from our 2014 paper, "Structured light self-calibration with vanishing points" (Dr. Radu Orghidan, EMBA, Joaquim Salvi, Mihaela Gordan, Camelia Florea, Joan Batlle — Machine Vision and Applications), showing how both a camera and a projector can be calibrated from the same vanishing-point geometry, no DLT pipelines, no iterative optimization. It walks through the full pipeline step by step: vanishing point detection, camera calibration, projector calibration with keystone rectification, and 3D scanning with De Bruijn color-coded stripes. Everything runs in the browser, everything is computed from real pinhole-model math. What still stands out to me is the elegance of the underlying idea: recovering 3D geometry from perspective convergence. That mathematical insight translates directly into engineering value with fewer constraints, faster setup, lower cost. What's new is how it came back to life. Using AI-assisted specification-driven development, I translated research code written over a decade ago into a usable web application in a fraction of the time it would have taken even a few months ago. For me, this is evidence of one of the most exciting shifts happening right now: modern AI tools don't just generate code, they can resurrect deep technical work and make it accessible again. The path from paper to product is becoming dramatically shorter, and a lot of valuable research is waiting to be reactivated. https://lnkd.in/esiKYiBT #ComputerVision #3DScanning #StructuredLight #Calibration #VanishingPoints #AIEngineering #WebApp #ResearchToProduct #AugmentedReality

Why Physics and Math are the hidden engines of Computer Science. ⚙️ As I dive deeper into my A-Level CS, Math, and Physics studies, I’ve realized that coding isn't just about syntax—it’s about understanding the physical and logical laws that govern our world. Whether it’s using mechanics to understand game engine physics or applying complex calculus to optimize algorithms, the crossover between these subjects is where the real innovation happens. I’m currently exploring how these foundations apply to AI, Software Architecture etc. It’s a challenging journey, but seeing the "why" behind the code makes it worth it. What was the one subject that changed how you view technology? #ALevels #ComputerScience #TechCommunity #FutureEngineers #STEM

🚀 Open-Sourcing a New Paradigm in Computational Chemistry: The Kyjovský Topological Engine For decades, material design has relied on solving complex Schrödinger equations, probabilistic fermion clouds, and orbital hybridizations. While this works, the computational cost (FLOPs) required to simulate complex molecules is massive. What if we bypass these heavy probability integrals entirely using pure wave geometry? I am thrilled to announce that the Python Minimum Viable Product (MVP) of the Kyjovský Topological Model is now live and open-source on GitHub! 💻 Instead of treating the vacuum as "empty space" and electrons as orbiting point-particles, this engine models a fully structured 55-point Higgs lattice (Mackay icosahedron). We treat molecular stability strictly as a topological problem. Here is how the engine recalculates physics: ⚛️ Nucleon Volume: 1 nucleon (proton or neutron) occupies exactly 3.25 vacuum clusters. 🌊 The Wave Electron: The electron is not a physical particle. It is a standing resonance wave generated by the outward-pointing vectors (gluons) of protons. 🔗 Phase Alignment: Chemical bonds are computed via simple scalar phase alignment using a "Cluster Index" (K_{cc}), rather than complex electron repulsion. Whether it’s mathematically proving the directional sp^3 valence nodes of Carbon as an overcrowded spatial layer, or predicting the absolute resonance of Gallium Nitride (GaN) in milliseconds, this heuristic algorithm reduces chemistry to elegant integer mathematics. If you are a computational chemist, a materials science engineer, or an AI developer looking for exponentially faster algorithmic heuristics for physical simulations, I invite you to explore the code. Clone the repo, run the simulations, and let's rethink physics from the ground up: 🔗 https://lnkd.in/dUxdKtjY Author: Peter Kyjovsky ORCID iD: 0009-0008-3806-1964 #ComputationalChemistry #MaterialsScience #QuantumPhysics #Python #OpenSource #Simulation #DeepTech #Innovation #Physics

Roger Penrose, the Nobel Prize winner of 2020. The author of the influential book The Road to Reality, has offered profound insights into the foundations and limitations of Einstein's equations. Tensor geometry provided Einstein general relativity with a language. Hereafter known as GR, yet quantum physics soon revealed its limitations. The smooth, continuous spacetime and local deterministic evolution of GR are fundamentally at odds with the discreteness, uncertainty, superposition, and entanglement inherent in quantum mechanics. As a result, GR remains incomplete when confronted with quantum phenomena. Einstein's General Relativity is the ultimate classical physics theory with gravity as curved spacetime, described by elegant tensor equations, but push it to quantum scales, and it catastrophically breaks down. Singularities produce infinities, and spacetime loses its smooth geometry. GR directly contradicts quantum mechanics most sacred law. That information can never be destroyed. A century after Einstein, we still have no experimentally verified theory of quantum gravity from any lab. Tensors enabled Einstein to encode gravity into the geometry of spacetime, but quantum theory soon revealed the cracks in this framework. GR treats spacetime as smooth, local, and classical, whereas quantum mechanics emphasizes Heisenberg’s principle, superposition, entanglement, and the possibility of quantized geometry. Quantum Field Theory (QFT) excels at describing the quantum behavior of non gravitational force, while Einstein’s field equations remain unmatched in the classical description of the gravity domain. However, Einstein’s equations are not equipped to withstand the demands of the quantum frontier without significant modification. Tensors let Einstein turn gravity into geometry. Roger Penrose helps show why that triumph is still not complete. GR assumes a smooth, classical spacetime, while quantum physics points to QFT, superposition, entanglement, and perhaps granular geometry itself. Einstein’s field equations were a masterpiece of classical physics but do not survive unchanged when applied to the quantum domain. #BigData #Analytics #DataScience #AI #MachineLearning #IoT #IIoT #PyTorch #Python #RStats #TensorFlow #Java #JavaScript #ReactJS #GoLang #CloudComputing #Serverless #DataScientist #Linux #Books #Programming #Coding #100DaysofCode https://lnkd.in/gsmnJUJe

⚛️ From Quantum Mechanics Foundations to Chip-Scale Maser Understanding quantum behavior in real devices is complex. Whether it's a two-level system (|g⟩ & |e⟩) in a diamond NV-center or population inversion in a maser — the physics starts from the same place: Quantum mechanics. But visualizing this in real-time? Almost impossible with static diagrams. 💡 Our Solution: My collaborator Shahariar Ryehan and I built a Python-powered animated simulation that walks through every key concept — from wavefunction to working maser. 🔑 Key Insights: → Two-Level System: Energy states |g⟩ (ground) in BLUE and |e⟩ (excited) in RED — the heart of every maser → Wavefunction Probability: dP = |ψ(r,t)|² dV — where the quantum particle is likely to be found → Density Matrix: ρ̂ = [[ρₑₑ, ρₑ₉], [ρ₉ₑ, ρ₉₉]] — handles mixed states in real experiments → Boltzmann Distribution: At room temperature (300K), Nₑ/N₉ ≈ 0.9995 — NO maser! → Population Inversion: Pump laser makes Nₑ > N₉ — maser condition achieved! → Chip Challenge: Surface defects cause T₁ relaxation — energy leaks out → The Solution: Fast optical pumping (532nm green laser) with P_pump > 1/T₁ → Final Result: When condition met → ρₑₑ > ρ₉₉ → MASER OSCILLATES! 🛠️ Built with: Python, Matplotlib, NumPy, FFMpeg 🔗 Run the code & watch the animation: https://lnkd.in/gpm6wtgF https://lnkd.in/g7yUc_EH Huge thanks to my collaborator Shahariar Ryehan for numerical implementation & animation design. #QuantumMechanics #Maser #DiamondNV #PopulationInversion #Python #Simulation #TwoLevelSystem #DensityMatrix #BoltzmannDistribution #ChipScaleMaser

I am thrilled to announce that our (Yannis Psaromiligkos) paper, "𝐌𝐨𝐝𝐞𝐥-𝐛𝐚𝐬𝐞𝐝 𝐃𝐞𝐞𝐩 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐟𝐨𝐫 𝐉𝐨𝐢𝐧𝐭 𝐑𝐈𝐒 𝐏𝐡𝐚𝐬𝐞 𝐒𝐡𝐢𝐟𝐭 𝐂𝐨𝐦𝐩𝐫𝐞𝐬𝐬𝐢𝐨𝐧 𝐚𝐧𝐝 𝐖𝐌𝐌𝐒𝐄 𝐁𝐞𝐚𝐦𝐟𝐨𝐫𝐦𝐢𝐧𝐠" has been accepted for publication in the journal: 𝐈𝐄𝐄𝐄 𝐖𝐢𝐫𝐞𝐥𝐞𝐬𝐬 𝐂𝐨𝐦𝐦𝐮𝐧𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬 𝐋𝐞𝐭𝐭𝐞𝐫𝐬 𝐈𝐄𝐄𝐄 𝐗𝐩𝐥𝐨𝐫𝐞: https://lnkd.in/ewPErAZw 𝐚𝐫𝐗𝐢𝐯: https://lnkd.in/eahUiJyh 𝐃𝐎𝐈: 10.1109/LWC.2026.3683016 𝐂𝐨𝐝𝐞: https://lnkd.in/eZa_5M5s 𝐀𝐛𝐬𝐭𝐫𝐚𝐜𝐭:  A model-based deep learning (DL) architecture is proposed for reconfigurable intelligent surface (RIS)-assisted multi-user communications to reduce the number of bits required for transmitting phase shift information from the access point (AP) to the RIS controller. The AP computes the phase shifts and compresses them into a binary control message that is sent to the RIS controller for element configuration. To help reduce beamformer mismatches caused by phase shift compression errors, the beamformer is updated with the actual (decompressed) RIS phase shifts. By unrolling the iterative weighted minimum mean square error (WMMSE) algorithm within the wireless communication-informed DL architecture, joint phase shift compression and WMMSE beamforming can be trained end-to-end. Simulation results demonstrate that incorporating compression-aware beamforming significantly improves sum-rate performance, even when the number of control bits is lower than the number of RIS elements. #DeepLearning #WirelessCommunications #RIS #IEEE #MachineLearning #SignalProcessing