Scientific Computing Software Tools

I spent the last few days rebuilding 20 of the figure types frequently appear in Nature and Cell papers, all in R with ggplot2. Manhattan and volcano plots, Circos diagrams, Sankey flows, raincloud and split-violin plots, treemaps, Mantel heatmaps, and more. Each figure comes with its own data simulator, so the whole thing runs on a fresh clone with nothing to download. Change one parameter and the figure updates with it. Drop in your own data when you're ready. I built it to get comfortable with publication-quality plotting for my genomics work, and to save the next person some of the trial and error. Github repo link for code and all 20 examples: https://lnkd.in/gHvgbr9X Acknowledgement: The idea for these figures came from a WeChat post shared by Dr. Rana Muhammad Atif. I then built the whole thing my own way, as a clean and freely accessible repository. #RStats #ggplot2 #Bioinformatics #DataVisualization #PlantScience #DataScience #Genomics #RStats #ComputationalBiology #PlantScience #Rprogramming

𝗠𝗮𝘁𝗹𝗮𝗯, 𝗣𝘆𝘁𝗵𝗼𝗻, 𝗼𝗿 𝗝𝘂𝗹𝗶𝗮 — 𝘄𝗵𝗶𝗰𝗵 𝗶𝘀 𝗳𝗮𝘀𝘁𝗲𝘀𝘁 𝗳𝗼𝗿 𝘀𝗰𝗶𝗲𝗻𝘁𝗶𝗳𝗶𝗰 𝗰𝗼𝗺𝗽𝘂𝘁𝗶𝗻𝗴? I tried answering this question about 10 years ago while working on dynamic Finite Element (FE) simulations. My scripts back then were largely memory-bound: big FE models with relatively simple work per byte transfered. Not surprisingly, speed differences between the three languages were small.    Benchmarking programming languages for speed isn’t trivial. The outcome is problem-dependent (e.g. memory-bound vs. compute-bound) and literal translations from one language to another rarely make for a fair comparison. Also, the more sophisticated the benchmark, the more it reflects programming skill, not language speed. While AI coding assistants have helped programmers level-up almost overnight, I’m still a far cry from a programming wizard. Nevertheless, the revived Julia momentum tempted me to a new benchmark.   I picked two simulations that are rich in physics yet fairly simple to implement: • Fluid flow past a cylinder using the lattice Boltzmann method: largely memory-bound; representative of many lower-order PDE codes;   • Two colliding galaxies (all-pairs N-body): largely compute-bound; think moderate-size lumped-parameter models with expensive right-hand sides (e.g. nonlinear tyre forces). For each case I started with a baseline Python implementation (the kind PhD students in engineering or sciences come up with when speed is of no concern), then literally translated it to Matlab and Julia. After the baseline benchmark, I optimized each script in its own idiom (to the best of my abilities and obviously AI-assisted) and ran a second analysis. Note: for the N-body case, optimizing the Matlab and Python codes didn’t yield meaningful speedups. I’ve made all scripts used for this benchmark available at https://lnkd.in/erpcFk4R, so you can rerun, critique, improve, or use it to benchmark your own scripts.     For compute-heavy simulations, Julia does seem to live up to its high performance reputation, running about 4-5 times faster than Matlab and Python in the N-body case. For memory-bound simulations, the efficiency gain is much less – but given Julia’s current allround maturity, that’s hardly a show-stopper. Want to know whether your Matlab or Python script can benefit from conversion to Julia? If you’re not into counting FLOPs & bytes, a simple experiment is to add "useless" FLOPs to your kernels and monitor how that influences runtime: if it grows proportionally, your code is compute-bound and Julia may be worth a try. 

🚀 Computational Physics: The Power of Science and Computing Computational Physics integrates physics, applied mathematics, and computer science to solve complex problems through numerical simulations. For professionals in modeling, machine learning, or data science, mastering this field unlocks groundbreaking applications. --- 🔍 What is Computational Physics? Many physical problems lack exact solutions, requiring numerical methods to obtain approximate answers. Computational Physics enables us to: ✔ Solve differential equations in complex systems. ✔ Simulate phenomena that can't be tested in labs. ✔ Model chaotic systems like climate and turbulence. ✔ Process and interpret vast scientific data. --- ⚙ Key Methods and Techniques 📌 Finite Difference (FDM), Finite Element (FEM), and Finite Volume (FVM) – Essential for engineering, fluid dynamics, and electromagnetism. 📌 Molecular Dynamics (MD) & Monte Carlo (MC) – Used in biophysics and materials science. 📌 Numerical Linear Algebra – LU decomposition, FFT, and conjugate gradient methods for efficient computations. --- 🌍 Applications of Computational Physics 🔭 Astrophysics & Cosmology – Simulating black holes, galaxy evolution and finding new exoplanets. ⚛ Particle Physics & Quantum Mechanics – Modeling high-energy collisions. 🌪 Computational Fluid Dynamics (CFD) – Applied in weather forecasting and aerodynamics. 🔬 Materials Science – Simulating semiconductors, nanotechnology, and superconductors. 🤖 Machine Learning & Physics – Neural Networks for solving PDEs and accelerating simulations. --- 🛠 Essential Tools & Programming Languages 💻 Languages – Python (NumPy, SciPy), C/C++, Fortran, Julia. 📊 Software & Frameworks – MATLAB, COMSOL, OpenFOAM, LAMMPS, GROMACS, ROOT. --- 🚀 Future Trends & Challenges ⚡ High-Performance Computing (HPC) – Enabling precise simulations. 🧠 AI in Physics – Implementing Physics-Informed Neural Networks (PINNs). 💡 Quantum Computing – Potential breakthroughs in quantum mechanics. --- 🔗 Conclusion From exploring the universe to predicting climate patterns and designing materials, Computational Physics is a powerful tool shaping science and technology. Mastering scientific programming and numerical modeling is an excellent starting point. 💬 Have you worked with Computational Physics before? Or interested in learning more? Let’s discuss in the comments! 👇

Imagine we could map every cell in the human body, revealing its precise location and molecular identity. This tantalizing possibility is at the heart of our latest perspective piece published in Nature Methods, where we explore a groundbreaking approach to understanding biological systems at unprecedented depth and scale: Deep 3D Histology. In this perspective article, we discuss three key pillars of this emerging field: Advanced Tissue Clearing and Imaging: -Cutting-edge tissue clearing techniques for intact specimen visualization -High-resolution light-sheet microscopy pushing the boundaries of 3D imaging -Applications ranging from mouse embryos to entire human organs Spatial Omics Technologies: -Integration of single-cell omics data with 3D spatial context -Creation of comprehensive molecular atlases of entire organisms -Bridging the gap between molecular profiles and tissue architecture Artificial Intelligence in Image Analysis: -Deep learning revolutionizing 3D histology data processing -Automation of tasks from image enhancement to cell segmentation -Unveiling information invisible to the human eye through "virtual staining" The potential impact of combining these technologies is staggering. By accelerating our understanding of diseases and drug discovery, we could compress centuries of insights into just a few years of research. Challenges remain, including improving resolution, increasing imaging speed for large samples, and developing user-friendly AI tools. But as we overcome these hurdles, Deep 3D Histology could become a routine tool in both research and clinical settings. The future of biomedical research is three-dimensional, molecularly detailed, and AI-enhanced. This new era of 3D omics has the potential to revolutionize medicine and our understanding of life itself. You can read the full perspective and join the discussion on this exciting frontier of science: https://rdcu.be/dNBe8 More technical details are here as tweetorial: https://lnkd.in/d48cTXDE #AI #DeepLearning #Clearing #3D #Imaging #Omics #Deep3DHistology

In Tao Yu's recent paper published in Computers and Geotechnics, we’ve developed a groundbreaking GPU-accelerated framework that significantly speeds up fracture simulations using the phase-field method (PFM) within a finite volume method (FVM) framework. By leveraging the power of GPU parallelism, our approach achieves up to 12x speedup compared to traditional CPU-based methods and enables large-scale, high-resolution simulations previously deemed computationally infeasible. The new framework features full GPU parallelization from mesh processing to linear solver and is validated across multiple benchmark cases (e.g., single-edge notched tension, L-shaped panel). It can be seamlessly coupled with two-phase CFD for hydraulic fracturing simulations. With a single GPU, it may reaches up to 120x equivalent CPU-core performance. This opens new doors for high-fidelity modeling of complex fracture processes in geotechnical engineering, reservoir simulation, and beyond with reduced computational cost and time. Please read the full open-access article here: DOI: 10.1016/j.compgeo.2025.107481 or Preprint from: https://lnkd.in/g6sxqVYV #GPUComputing #PhaseField #FractureMechanics #ComputationalGeomechanics #FiniteVolumeMethod #HPC #Research #OpenAccess #Geotech #CFD #HydraulicFracturing #HKUST

Sometimes university exams force you to go beyond just writing definitions and actually understand how things work under the hood. I spent time breaking down CUDA Architecture from scratch and visualizing the complete execution flow. What I learned: • How the Host (CPU) and Device (GPU) communicate through PCIe. • How data moves from Host Memory to Device Memory. • How a CUDA Kernel is launched from the CPU and executed on the GPU. • The hierarchy inside a CUDA program: Grid → Blocks → Threads. • The internal structure of a GPU and Streaming Multiprocessors (SMs). • The role of CUDA Cores, Warp Schedulers, Registers, Shared Memory, L1/L2 Cache, and Global Memory. • How thousands of threads execute simultaneously to achieve massive parallelism. Understanding CUDA is not just important for exams. It is the foundation behind many modern computing applications: ✓ Deep Learning & AI Training ✓ Large Language Models (LLMs) ✓ Computer Vision ✓ Scientific Simulations ✓ Autonomous Vehicles ✓ Video Rendering ✓ High-Performance Computing (HPC) ✓ Data Analytics and Big Data Processing A simple real-world example: When training a neural network with millions of images, a CPU would process tasks sequentially and take significantly longer. CUDA allows the GPU to distribute computations across thousands of cores and execute them in parallel, reducing training time from days to hours. The biggest takeaway: CPU = Good for complex control logic. GPU + CUDA = Excellent for performing the same operation on massive amounts of data simultaneously. #CUDA #GPUComputing #ParallelComputing #NVIDIA #MachineLearning #DeepLearning #ComputerEngineering #AI #DataScience #UniversityLearning

Tried something interesting with automated annotations on histology images. Using a simple text prompt, Grounding DINO was able to detect glomeruli in a kidney H&E image and generate bounding boxes. I then passed those boxes to SAM 2, which converted them into clean, pixel-level segmentation masks. So essentially: text prompt → object detection → precise segmentation. What stood out to me is how well this worked on scientific imagery, not just natural images. Annotation is often a bottleneck in biomedical workflows, and this kind of pipeline could significantly speed things up. Would be interesting to see how robust this is across different stains, tissues, and imaging conditions. If there’s interest, I can put together a short video walkthrough. #microscopy #computervision #deeplearning #imageanalysis #digitalpathology

One of the biggest performance gains in electromagnetic transient (#EMT) studies doesn't come from buying a faster computer (though that helps!). It comes from running simulations in parallel. This is by far the most common question I hear relating to simulation speed improvements. Whether you're using #PSCAD or any other #powersystem #simulation tool, parallel processing is essential for completing large studies in a practical amount of time. Without it, many engineering studies would simply take too long to be useful. The default PSCAD™ license supports up to 8 simultaneous #EMTDC instances, allowing multiple simulations to execute at the same time. In practice, I tend to see two common approaches. First, a single very large model can be divided into smaller sections along transmission lines. Each section runs as its own EMTDC instance, allowing the overall simulation to be distributed across multiple processor cores. This can significantly reduce simulation time for large interconnected networks. Second, and perhaps even more common, is running many independent cases simultaneously. Individual simulations may be relatively small, but studies often require hundreds or even thousands of cases. Examples include MQT studies, PRC 029 assessments, and large parameter sweeps. Instead of running these sequentially, the cases are distributed across the available EMTDC instances, dramatically reducing the total study time. This ability to parallelize simulations has become a critical part of modern power system engineering workflows. As networks become more complex and study requirements continue to grow, efficiently utilizing available computing resources is no longer just a convenience. It is a necessity. If there is interest I can write a post about how and why transmission lines are used to split cases for parallel processing. Let me know!

K-space is the heart of MRI physics—yet it remains one of the most misunderstood concepts in medical imaging. Rather than representing an image directly, k-space stores raw data in the frequency domain. Every point in k-space contains information about the entire image, and the final MR image is reconstructed using a mathematical process called the Fourier Transform. The center of k-space holds contrast and overall signal intensity, while the outer regions define image detail and spatial resolution. This is why motion or errors during acquisition can significantly affect image quality—especially when the central lines are disrupted. Understanding k-space is crucial for radiographers when optimizing sequences, managing artifacts, and improving scan efficiency. Techniques like parallel imaging, partial Fourier, and compressed sensing all rely on intelligent k-space sampling. In MRI, mastering k-space is not just physics—it’s the key to producing high-quality, diagnostic images. #MRIPhysics #KSpace #Radiography #MedicalImaging #FourierTransform #MRI #ImagingScience #ContinuousLearning

🔴 Researchers from imec, EPFL, KTH Royal Institute of Technology, and Tyndall National Institute present the blueprint for next-generation integrated photonics in #MicrosystemsAndNanoengineering. The paper "Integrated silicon photonic MEMS" proves that combining micro electromechanical systems with standard foundry processes will define the next decade of #SiliconPhotonics and #OpticalComputing. While silicon photonics has emerged as a mature technology for high data rate communications and autonomous vehicle sensing, the material's weak electro-optic effects remain a bottleneck. Traditional thermo-optic tuning devices demand continuous power consumption and result in large footprints. This comprehensive research proves that integrating MEMS directly into silicon photonic circuits is the ultimate solution. 1️⃣ Overcoming Material Limits: #PhotonicMEMS & #EnergyEfficiency By replacing bulky traditional modulators with silicon photonic MEMS, the architecture drastically reduces the device footprint. Furthermore, it introduces bistable phase switches that enable nonvolatile photonic circuits, eliminating the need for continuous static power consumption. 2️⃣ Wafer Level Scalability: #Foundry & #Packaging The true breakthrough lies in manufacturability. The research highlights the successful implementation of wafer-level hermetically sealed packaging. This ensures that these advanced MEMS components can be produced with high yield and high volume capacity using standardized silicon foundries. 3️⃣ Reconfigurable Architectures: #OpticalRouting & #QuantumInformation This scalable integration provides access to fully reconfigurable coupled resonator optical waveguides. It unlocks optimized library components for complex optical routing, paving the way for advanced photonic accelerated computing and quantum information processing. 💡 My Take: As the demands of AI and data centers push optical communication to its limits, the massive power consumption of thermo-optic tuning in traditional silicon photonics is no longer sustainable. By physically moving microscopic silicon structures using MEMS, we can route light with near-zero static power. This research is a massive wake-up call for the industry. Transitioning from solid-state thermal tuning to wafer-scale integrated photonic MEMS is not just an incremental hardware update, it is a mandatory architectural revolution required to build energy-efficient, large-scale optical networks. 👇 Link in the comments #AdvancedPackaging #HardwareArchitecture #Metrology #3DIC #DataCenter #AIHardware #Telecommunications #Optoelectronics Intel TSMC Samsung Electronics GlobalFoundries NVIDIA Broadcom Marvell Technology Cisco Applied Materials ASML Lam Research Lumentum Coherent Corp. Infinera STMicroelectronics