𝗕𝗲𝘄𝗮𝗿𝗲 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗾𝘂𝗶𝗿𝗸 𝘄𝗵𝗲𝗻 𝗰𝗿𝗲𝗮𝘁𝗶𝗻𝗴 𝗮𝗴𝗲𝗻𝘁𝘀 𝗽𝗿𝗼𝗴𝗿𝗮𝗺𝗺𝗮𝘁𝗶𝗰𝗮𝗹𝗹𝘆 In this guest blog post by Augusto Pereira, we discuss why, when creating agents programmatically, you need to set parameters in the agent constructor, and why variables behave differently. Knowing about this feature when building a large AnyLogic model can save you hours of debugging when overlooked. https://lnkd.in/dBehBwqY #TheAnyLogicModeler #Simulation #AnyLogic

Sherlock STEAM — First Term (50+ Scenes) is live. I’ve bundled the entire run — Parts 1–50+ — into a single PDF for fast access and teaching/use in teams. Each scene is a micro-lesson with "locked-axis visuals", "hardware-safe practices", and "Sora 2 / WAN 2.5 / Veo 3.1"-ready prompts. What’s inside * Foundations: On/Off with hysteresis; P, I, D; PD/PI; "PID tuning" (gentle first → Z-N). * Protect the hardware: Setpoint ramps & rate limits, anti-windup, scheduling & anti-chatter, safety interlocks. * Smarter loops: Feed-forward , disturbance observer (DOB), "noise filtering"(median vs LPF), "reference pre-filters". * Better sensing & models: "Sensor fusion" (IMU+Baro), observers (Luenberger/Kalman), "LQR", "MPC", reference governors. * Advanced & resilient: "Robustness badges", bumpless transfer, "fault-tolerant" control, energy-aware MPC, networked control, "secure control". * Learning & adaptation: "Adaptive PID", fuzzy hybrids, "learning-based control" (safe RL warm-start), meta-learning, "auto-calibration", A/B on a Digital Twin. * Finale: Mini and extended supercuts with clean summaries and CTAs. Why it helps * Consistent format: Tagline • What it teaches • Key visuals • Safety/Style. * Immediate reuse: Short, readable labels; identical axes for fair comparisons. * Education & R&D-friendly: Perfect for stand-ups, onboarding, workshops, and course modules. #Steam #Stem #SteamEducation #EngineeringEducation #SherlockTeam #ControlSystems #PID #MPC #Robotics #Automation #SignalProcessing #DigitalTwin #RobustControl #SensorFusion #SafeByDesign #Sora2 #wan25 #veo31 #KamilBala

YOLOv8 represents a significant evolution in real-time object detection, balancing speed and accuracy remarkably well. What stands out to me is how it simplified the API compared to earlier YOLO versions-making it more accessible through the Ultralytics package while maintaining the performance improvements. The anchor-free detection approach was a smart move, reducing hyperparameters and making it more generalisable across different datasets. It's also interesting how YOLOv8 isn't just about object detection anymore-it extends to segmentation, pose estimation, and classification tasks, making it a versatile framework for various computer vision applications. For practical deployment, especially on edge devices or in applications requiring real-time inference, YOLOv8 hits a sweet spot that few other models achieve.

This is a cool technology! I’ve used KinExA technology for close to three decades — measuring affinities in the low femtomolar (fM) range and also using it to screen for very high-affinity antibodies. For example, in the following paper, I measured affinities in the ~34 fM range and also used KinExA for high-affinity antibody screening: 📄 Development of a potent high-affinity human therapeutic antibody via novel application of recombination signal sequence-based affinity maturation. Kielczewska A, D'Angelo I, Amador MS, Wang T, Sudom A, Min X, Rathanaswami P, Pigott C, Foltz IN. J Biol Chem. 2022 Feb;298(2):101533. [PMID: 34973336] In the following earlier work, we demonstrated how to measure femtomolar-range affinity interactions: 📄 Demonstration of an in vivo generated sub-picomolar affinity fully human monoclonal antibody to interleukin-8. Rathanaswami P, Roalstad S, Roskos L, Su QJ, Lackie S, Babcook J. Biochem Biophys Res Commun. 2005 Sep 9;334(4):1004–1013. [PMID: 16038881] It’s wonderful to see this technology continue to evolve and thrive after all these years, and I truly appreciate that Sapidyne has now developed a methodology to apply this powerful technology right from the screening stage. 🔬

🚀 Fixstars Amplify AE v1.0 is now live! https://lnkd.in/gMWf5iRD The latest update to our GPU-based Ising machine delivers faster, easier, and more scalable solutions for real-world combinatorial optimization. ✨ Highlights • Boosted GPU performance for large-scale problems • Automated parameter tuning — no manual trial and error • Support for inequality constraints & higher-order polynomials From manufacturing to finance, Amplify AE v1.0 helps unlock practical optimization at scale.

#AutoTwinNews 📢 𝗔 𝗗𝗲𝗲𝗽 𝗗𝗶𝘃𝗲 𝗶𝗻𝘁𝗼 𝘁𝗵𝗲 𝗔𝘂𝘁𝗼-𝗧𝘄𝗶𝗻 𝘁𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗲𝘅𝗽𝗹𝗮𝗶𝗻𝗶𝗻𝗴 𝗦𝘆𝘀𝘁𝗲𝗺 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗚𝗿𝗮𝗽𝗵𝘀. In our latest video, we walk you through how AUTO-TWIN Project automates the process of: ⚙️ Integrating and enriching data with domain knowledge 🧩 Fixing data quality issues 🔄 Modeling process executions and system behavior 🤖 And orchestrating it all through the Auto-Twin Orchestrator Watch the video 🎬 https://lnkd.in/d-32GVJi narrated by Dirk Fahland (Eindhoven University of Technology) Learn more 💡 https://lnkd.in/dXMdkqX5 #SKG #digitaltwins #circulareconomy #horizoneu

We are thrilled to announce a strategic partnership with EdgeTech Marine! We are launching #SeaView #eBOSS, a powerful new module for our software platform.  Coming soon, SeaView eBOSS will provide full, end-to-end processing and interpretation capabilities for Edgetech's industry-leading eBOSS instrument. SeaView eBOSS allows users to seamlessly integrated eBOSS volue data with #SSS, #MBES, #SBP and #MAG datasets for a comprehensive, multi-modal analysis.  It can track boulders and cables, estimate dimensions and volumes, and automatically compute burial depths, making it the ideal software platform for #routesurveys, #uxo, #cablelay and #subsea #asset monitoring projects. Want to be the first to know when SeaView eBOSS launches? Sign up for updates here: https://lnkd.in/dNMxnMwP

📚 Learn how to deploy vLLM on NVIDIA DGX Spark — the right way! NVIDIA AI just published a detailed best practices guide for running high-throughput inference with vLLM, including multi-node setups and optimized Docker builds. 👉 Dive in: https://lnkd.in/e6tRirwr #vLLM #NVIDIA #DGXSpark #LLMInference #AIInfrastructure

💡In the new version 25.11 of HALCON, which was released on November 12, 2025, the software will be expanded with additional deep learning features. The focus is on speed. “Speed is a crucial factor in ensuring that deep-learning-based applications can be used profitably in practice Features: ✅ Continual Learning – Classification ✅ Score Visualization for Shape Matching ✅ Optimized Deep OCR models for faster, resource-efficient OCR ✅ MobileNetV4 classification models ✅ Various code reading and print quality 💡 This is a breakthrough Technology of MVTec, bringing high efficiency and accuracy to the Automation industry. Contact RTC Technology - MVTec's strategic partner in Vietnam to learn about Halcon software applications in manufacturing industries.

After experimenting with SAM2 for plant segmentation, I decided to take things one step further and train my own model. Using Roboflow and YOLOv11-seg, I annotated about 100 images of plants, added some augmentations, and trained a custom instance segmentation model directly in Google Colab - following this great tutorial: https://lnkd.in/d-jBBcrv. Within minutes I had a fully fine-tuned YOLOv11 model on my own data - and the results are fantastic. Next, I'll run inference on a few hundred new RGB images, use the resulting masks to isolate the same regions in my aligned thermal images, and start extracting agricultural insights by combining the visual and thermal data. It's amazing how accessible this workflow has become - from annotation to a production-ready segmentation model, all in a single Colab notebook. #ComputerVision #AgTech