Samer P. Francy, MSc, CSE, PMP’s Post

Check out the second video in our series, where we guide you through setting up multiple fatigue analyses and tasks in the frequency domain using FATIQ v25.x! We've got everything covered from step-by-step setup to interactive post-processing features. Watch now: https://lnkd.in/dgMHnG5E #FATIQ #FATIQv25 #FatigueAnalysis #BETACAE #PhysicsOnScreen

Day 11 of the #DrGViswanathan Challenge! 🚀 Today's focus: the classic 3Sum problem. A perfect example of using efficient algorithms to find unique triplets that sum to zero. Tackling challenges like these strengthens problem-solving skills and deepens my understanding of array manipulation. 💻🔍 #DSA #3SUM #ProblemSolving #Day11 #GrowthMindset

Self-distillation methods using Siamese networks are popular for self-supervised pre-training. DINO is one such method based on a cross-entropy loss between $K$-dimensional probability vectors, obtained by applying a softmax function to the dot product between representations and learnt prototypes. Given the fact that the learned representations are $L^2$-normalized, we show that DINO and its derivatives, such as iBOT, can be interpreted as a mixture model of von Mises-Fisher components. With this interpretation, DINO assumes equal precision for all components when the prototypes are also $L^2$-normalized. Using this insight we propose DINO-vMF, that adds appropriate normalization constants when computing the cluster assignment probabilities. Unlike DINO, DINO-vMF is stable also for the larger ViT-Base model with unnormalized prototypes. We show that the added flexibility of the mixture model is beneficial in terms of better image representations. The DINO-vMF pre-trained model consistently performs better than DINO on a range of downstream tasks. We obtain similar improvements for iBOT-vMF vs iBOT and thereby show the relevance of our proposed modification also for other methods derived from DINO. #selfdistillation #siamesenetworks #selfsupervisedlearning #DINO #vonMisesFisher #imageRepresentations #pretraining #IBOT #mixturemodel

Self-distillation methods using Siamese networks are popular forself-supervised pre-training. DINO is one such method based on a cross-entropyloss between K-dimensional probability vectors, obtained by applying asoftmax function to the dot product between representations and learntprototypes. Given the fact that the learned representations areL^2-normalized, we show that DINO and its derivatives, such as iBOT, can beinterpreted as a mixture model of von Mises-Fisher components. With thisinterpretation, DINO assumes equal precision for all components when theprototypes are also L^2-normalized. Using this insight we propose DINO-vMF,that adds appropriate normalization constants when computing the clusterassignment probabilities. Unlike DINO, DINO-vMF is stable also for the largerViT-Base model with unnormalized prototypes. We show that the added flexibilityof the mixture model is beneficial in terms of better image representations.The DINO-vMF pre-trained model consistently performs better than DINO on arange of downstream tasks. We obtain similar improvements for iBOT-vMF vs iBOTand thereby show the relevance of our proposed modification also for othermethods derived from DINO. #selfdistillation #siamesenetworks #selfsupervisedlearning #DINO #vonMisesFisher #imageRepresentations #pretraining #IBOT #mixturemodel

One of the best parts of using Cyft? You don’t have to speak in any specific format—Just brain dump. Just talk like you normally would, and Cyft automatically formats everything into the PSA—effortlessly. It’s built to work the way you do, making it easier than ever to stay on top of your tasks without the hassle. Michael Copeland (Mikey) - Verified User #MSPCommunity #UseCyft #FromLastPlaceToLeaderBoard #TechUtilization #HighLevelTasksNow

Self-distillation methods using Siamese networks are popular forself-supervised pre-training. DINO is one such method based on a cross-entropyloss between K-dimensional probability vectors, obtained by applying asoftmax function to the dot product between representations and learntprototypes. Given the fact that the learned representations areL^2-normalized, we show that DINO and its derivatives, such as iBOT, can beinterpreted as a mixture model of von Mises-Fisher components. With thisinterpretation, DINO assumes equal precision for all components when theprototypes are also L^2-normalized. Using this insight we propose DINO-vMF,that adds appropriate normalization constants when computing the clusterassignment probabilities. Unlike DINO, DINO-vMF is stable also for the largerViT-Base model with unnormalized prototypes. We show that the added flexibilityof the mixture model is beneficial in terms of better image representations.The DINO-vMF pre-trained model consistently performs better than DINO on arange of downstream tasks. We obtain similar improvements for iBOT-vMF vs iBOTand thereby show the relevance of our proposed modification also for othermethods derived from DINO. #selfdistillation #siamesenetworks #selfsupervisedlearning #DINO #vonMisesFisher #imageRepresentations #pretraining #IBOT #mixturemodel

Self-distillation methods using Siamese networks are popular forself-supervised pre-training. DINO is one such method based on a cross-entropyloss between K-dimensional probability vectors, obtained by applying asoftmax function to the dot product between representations and learntprototypes. Given the fact that the learned representations areL^2-normalized, we show that DINO and its derivatives, such as iBOT, can beinterpreted as a mixture model of von Mises-Fisher components. With thisinterpretation, DINO assumes equal precision for all components when theprototypes are also L^2-normalized. Using this insight we propose DINO-vMF,that adds appropriate normalization constants when computing the clusterassignment probabilities. Unlike DINO, DINO-vMF is stable also for the largerViT-Base model with unnormalized prototypes. We show that the added flexibilityof the mixture model is beneficial in terms of better image representations.The DINO-vMF pre-trained model consistently performs better than DINO on arange of downstream tasks. We obtain similar improvements for iBOT-vMF vs iBOTand thereby show the relevance of our proposed modification also for othermethods derived from DINO. #SelfDistillation #SiameseNetworks #DINO #PreTraining #MixtureModel

Self-distillation methods using Siamese networks are popular forself-supervised pre-training. DINO is one such method based on a cross-entropyloss between K-dimensional probability vectors, obtained by applying asoftmax function to the dot product between representations and learntprototypes. Given the fact that the learned representations areL^2-normalized, we show that DINO and its derivatives, such as iBOT, can beinterpreted as a mixture model of von Mises-Fisher components. With thisinterpretation, DINO assumes equal precision for all components when theprototypes are also L^2-normalized. Using this insight we propose DINO-vMF,that adds appropriate normalization constants when computing the clusterassignment probabilities. Unlike DINO, DINO-vMF is stable also for the largerViT-Base model with unnormalized prototypes. We show that the added flexibilityof the mixture model is beneficial in terms of better image representations.The DINO-vMF pre-trained model consistently performs better than DINO on arange of downstream tasks. We obtain similar improvements for iBOT-vMF vs iBOTand thereby show the relevance of our proposed modification also for othermethods derived from DINO. #SelfDistillation #SiameseNetworks #DINO #PreTraining #MixtureModel

Prodigy infotech ML task 3 - This task involves using a pre-trained VGG16 model to extract features from images of cats and dogs. These features are then used to train a Support Vector Machine (SVM) classifier to distinguish between the two classes. The process includes data preprocessing, feature extraction, model training, evaluation, and prediction. hashtag #ProdigyInfoTech #MachineLearning #SVM

🚀 New Tutorial Alert: Displaying LIGGGHTS Particles with Realistic Size & Extracting Data in Paraview 🚀 I’m thrilled to share my latest tutorial, where I walk you through the process of visualizing particle simulations generated in the open-source LIGGGHTS software. Often, when particle data is loaded into Paraview, all particles are shown as uniform points, missing their actual sizes and dimensions. In this step-by-step tutorial, I cover: How to visualize particles in their true size and shape in Paraview for realistic 3D representation. How to extract critical data such as radius, coordinates, and velocity directly from Paraview. Practical tips to enhance your particle simulation and visualization workflow. This guide is ideal for engineers, researchers, and simulation enthusiasts looking to elevate their skills in particle simulations and data extraction. Whether you’re new to LIGGGHTS or looking to optimize your workflows, this tutorial provides practical and actionable insights. Check it out, and let’s connect if you’d like to dive deeper into LIGGGHTS, Paraview, or particle simulation techniques! 🌟 Tutorial link: https://lnkd.in/gUSz8EU2. #LIGGGHTS #Paraview #ParticleVisualization #SimulationTools #EngineeringSimulation #ScientificComputing #DataExtraction #PhysicsSimulation #3DVisualization #EngineeringTips #ResearchTools #SimulationWorkflow