Corné V.’s Post

📣 New research paper, supported by #Sec4AI4Sec, published in the Journal of Systems and Software 📣 ✒ "Addressing combinatorial experiments and scarcity of subjects by provably orthogonal and crossover experimental designs". 🔍 In Sec4AI4Sec, we deal with some validations with humans to understand how AI methods really help in real life. While doing these experiments with limited numbers of developers/ students, we need a way to design the experiments so that they would still have a significant result. This paper discusses 3 possible balanced designs for this kind of experiment: full factorial design, orthogonal balanced design (taguchi), and crossover balanced design (NEW!). ➡ Find out more: https://lnkd.in/d2Rr6dRd #DesignofExperiments #CrossoverExperimentalDesign #FullFactorialDesign #OrthogonalDesign Fabio Massacci Aurora Papotti Ranindya Paramitha

A new research #paper "Addressing combinatorial experiments and scarcity of subjects by provably orthogonal and crossover experimental designs", supported by Sec4AI4Sec project, was published in the Journal of Systems and Software. This paper, written by Fabio Massacci, Aurora Papotti, Ranindya Paramitha, discusses 3 possible balanced designs for this kind of #experiment: full factorial design, orthogonal balanced design, and crossover balanced design. Find out more: https://lnkd.in/d2Rr6dRd #DesignofExperiments #CrossoverExperimentalDesign #FullFactorialDesign #OrthogonalDesign

🚨 Latest group preprint: 🚨 𝐅𝐞𝐍𝐍𝐨𝐥: 𝐚𝐧 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭 𝐚𝐧𝐝 𝐅𝐥𝐞𝐱𝐢𝐛𝐥𝐞 𝐋𝐢𝐛𝐫𝐚𝐫𝐲 𝐟𝐨𝐫 𝐁𝐮𝐢𝐥𝐝𝐢𝐧𝐠 𝐅𝐨𝐫𝐜𝐞-𝐟𝐢𝐞𝐥𝐝-𝐞𝐧𝐡𝐚𝐧𝐜𝐞𝐝 𝐍𝐞𝐮𝐫𝐚𝐥 𝐍𝐞𝐭𝐰𝐨𝐫𝐤 𝐏𝐨𝐭𝐞𝐧𝐭𝐢𝐚𝐥𝐬. 👉 : https://lnkd.in/enz4vrcb A new #GPU-accelerated #opensource library for building, training and running force-field-enhanced neural network potentials. It provides a flexible and modular system for building hybrid models, allowing to easily combine state-of-the-art embeddings with ML-parameterized physical interaction terms without the need for explicit programming. FeNNol shrinks the performance gap between ML potentials and standard force-fields. It can be used standalone or via Deep-HP within Tinker-HP: heavy #HPC optimization is underway for multi-(nodes/GPUs) runs. Available at https://lnkd.in/e4MCDRp2 Great work by Thomas Plé, Olivier ADJOUA, Louis Lagardère Funding European Research Council (ERC) (project EMC2). Supercomputer time GENCI. #drugdesign #NeuralNetworks #GPU #supercomputing #HPC NVIDIA #machinelearning Sorbonne Université CNRS

Michael Canesche's paper, which describes the new kernel fusion algorithm used in the Cadence XNNC Tensor Compiler, has been accepted at the International Conference on Compiler Construction. Tensor compilers like XLA, TVM, and TensorRT operate on computational graphs, where vertices represent operations, and edges denote data flow between these operations. Operator fusion is an optimization technique that combines multiple operators into a single, more efficient operation. The paper "Fusion of Operators of Computational Graphs via Greedy Clustering: The XNNC Experience" introduces the operator fusion algorithm recently implemented in the Xtensa Neural Network Compiler (XNNC). XNNC is a toolchain designed for deploying machine learning models on Cadence's Tensilica processors. These edge-device processors are widely used in applications such as automotive systems, consumer electronics, communications, LiDAR, and radar technologies. First released in 2017 to complement Tensilica’s Vision 7 processors, XNNC has since evolved significantly. Now in version 3.0, its codebase spans hundreds of thousands of lines of C++ code. XNNC has been used to compile thousands of neural networks for a broad range of Xtensa architectures, and its design and implementation continue to advance, as this paper demonstrates. Read the paper: https://lnkd.in/dsQJtSkz #compilers #research #university #education #gradschool

What is neuromorphic computing? Neuromorphic computing, also known as neuromorphic engineering, is an approach to computing that mimics the way the human brain works. It entails designing hardware and software that simulate the neural and synaptic structures and functions of the brain to process information. https://lnkd.in/exVR-uq3.

When will this be production ready and challenge NVIDIA? And save the environment through less energy consumption? “In the new paper, the researchers suggest replacing the traditional 16-bit floating point weights used in Transformers with 3-bit ternary weights that can take one of three states: -1, 0 and +1. They also replace MatMul with additive operations that provide equally good results at much less computational costs. The models are composed of “BitLinear layers” that use ternary weights.” https://lnkd.in/eg5dTiiE

🌟 Thrilled to Announce Our Latest Project! 🌟 I am delighted to share a groundbreaking project that my colleague, Emilio Rodrigo Carreira Villalta, and I have recently completed. Leveraging the YOLO (You Only Look Once) model from Ultralytics, we have successfully developed an advanced system capable of recognizing hand gestures in the game of Rock-Paper-Scissors. This project entailed comprehensive research and meticulous model training to achieve high accuracy in gesture detection and classification. Here are some key aspects of our work: 🔍 Model Training and Optimization: We employed the YOLO architecture to train our model on a custom dataset of hand gestures, focusing on optimizing its accuracy and performance, as well as to understand and study which are the necessary files and structure that the Ultralytics models need for their learning. In order to be able to training the model, we used the computational power of an NVIDIA T4 GPU on Google Colab, getting an accuracy of around 97%. 📊 Performance Metrics: Utilizing robust evaluation metrics, we rigorously assessed the model's effectiveness plotting the graphs that the model auto-generates, ensuring reliable and precise recognition of gestures. 🎥 Inference and Application: The trained model was applied to real-time video and image frames, demonstrating its capability to accurately interpret and classify gestures in various scenarios. This endeavor has significantly enhanced our expertise in machine learning, particularly in the realms of computer vision and real-time gesture recognition. It underscores the versatility and power of YOLO models in practical applications. I hope I can share many more projects with my colleague. Github repository: https://lnkd.in/dHVgCGfp #MachineLearning #ComputerVision #YOLO #ArtificialIntelligence #RockPaperScissors #Ultralytics #DeepLearning #GestureRecognition

See you at ICCS 2024: INTERNATIONAL CONFERENCE ON COMPUTATIONAL SCIENCE! Our CTO, Krzysztof Rojek, and the team will present our latest AI advancements and share insights on leveraging machine and deep learning to optimize complex algorithms and simulations. Don't miss our session: "Unleashing the Potential of Mixed Precision in AI-Accelerated CFD Simulation on Intel® CPU/GPU Architectures" to learn more. #ICCS2024 #AI #MachineLearning #DeepLearning #CFD #intelarchitectures https://lnkd.in/gnvcG4r8 CFD Suite (AI-accelerated CFD) Cognitive Services for Industries

Exciting news from the NRL! Professor Kaushik Roy has released a compelling new video titled "Re-Thinking Computing with Neuro-Inspired Learning: Sensors, Algorithms & Hardware Architecture." Don’t miss the chance to explore the innovative concepts that are shaping the future of computing. Check it out on his NRL YouTube Channel! #computing #Innovation #purdueuniversity #purdueece

Feature engineering is more important than model selection. Yeah, I said it. Fight me. Think your fancy neural network is gonna save you from crappy data? Like putting lipstick on a pig, except the pig is your career and the lipstick is overpriced GPUs. Cold, hard truth time: 🔪 Bad features turn complex models into lobotomized goldfish 🔪 Good features make complex patterns obvious 🔪 Better features = faster training, quicker inference 🔪 Domain knowledge > algorithmic wankery 🔪 Thoughtful engineering = interpretable models (or is that for wimps who can't defend black boxes?) Next time you're tempted to cosplay the latest arXiv paper, try engineering some decent features instead. Not convinced? Share your feature engineering tales of glory or humiliation. Bonus points for public embarrassment or pink slips. #DataScience #FeatureEngineering