Manish Sainani 🤫’s Post

I'm so excited to share the new paper of the I-FENN series (Integrated Finite Element Neural Network) with non-local gradient damage propagation. So far, the feasibility of our framework was shown on a single load-increment basis, and two crucial questions were lingering: Can we model the entire load-history of damage propagation with I-FENN, and is it indeed faster than conventional FEM? Here, for the first time, we address both questions. Check the answers in the link below! Mostafa Mobasher Habiba Eldababy @Diab Abueidda

Exciting news! Check out our latest paper where we model the entire load-history of damage propagation with I-FENN (Integrated Finite Element Neural Network) and demonstrate computational savings in various aspects of the framework.

You are given a series of numbers. Series: 52, 52, 40, 40, 28, 28, ... Give a simple mathematical formula to predict this series infinitely LLM can't do simple logic as of now. God my job is safe 🙏 Google Gemini ❌ https://lnkd.in/gfkRrG22 Term number (n) is even: n / 2 * (-12) + 52 Term number (n) is odd: 52 Anthropic Claude ❌ a(n) = 52 + (n - 1) * (-12) a(n) = 52 - 12n + 12 Therefore, the simple mathematical formula to predict this series infinitely is: a(n) = 64 - 12n Using this formula, we can generate the terms of the series as follows: a(1) = 64 - 12(1) = 52 a(2) = 64 - 12(2) = 40 a(3) = 64 - 12(3) = 28 a(4) = 64 - 12(4) = 16 a(5) = 64 - 12(5) = 4 OpenAI GPT 4 😎 https://lnkd.in/gmsxX77n an=52−12⌊n-1/2⌋ * lower ceiling function Altman is doing something right to take my job 😱

I am long overdue with Graph Attention Network on my Graph Neural Network series! In this video, we walk through the steps in creating a Graph Attention Network with Graph Attention Convolution layers. I hope this video enhances your understanding of graph attention! https://lnkd.in/gpAmsrZj

Hello guys I completed the task of YOLO_DETECTION YOLO (You Only Look Once) is a state-of-the-art, real-time object detection system. Unlike traditional methods that involve applying a classifier to an image at multiple locations and scales, YOLO applies a single neural network to the full image. This network divides the image into regions and predicts bounding boxes and probabilities for each region, enabling fast and accurate object detection. #AIMER #AIMERSOCIETY #APSCHE #object_detection #yolo #detected_cars #video

Starting this new year I decided to share my ML/AI knowledge so that both I (from reviews and comments) and the community can benefit from it. As a first step, I decided to write a very basic notebook on affine quantization of Neural Networks. In this process, I found an interesting platform called Ploutos for sharing knowledge. Kudos to Riju Pahwa for creating an amazing platform for knowledge sharing! Here is the very simple notebook affine quantization: https://lnkd.in/gyxyUAeJ. In the coming weeks I will try to create simple tutorials and share what I know about ML and AI to you all via this platform. Please feel free to point out any mistakes I make so that I can refine myself further! The reason I chose Ploutos as a platform and not Google Colab because: 1. Notebook UI similar to Colab. 2. Ability to create and run distributed jobs: I am planning to create some tutorials on distributed training too! 3. Ability to create endpoints: This will be useful to demo ML deployments. 4. Ability to create and run experiments. For example, large scale HParam tuning. 5. A nice way to organize things. and much more... Stay Tuned!

There’s an interesting, growing debate in the ML community on whether SORA is just a better video generator MODEL, or if it’s a new kind of « SIMULATOR », using a lot of synthetic data & learning its own model of physics (like the fluid dynamics example below). Letting AI learn the fundamentals of physics (intuitive physics) using gradient descent is a very interesting approach, next step in deep learning. Could the same approach be used to learn the fundamentals of cognition? The «e=mc^2» fundamental laws of our brain and decision making? Maybe the idea is to keep « bruteforcing » AI with # of parameters and scale to get to the point when we discover those fundamentals and simplify our models back using those fundamental laws… I guess we’ll see when Sama raises the $7 trillion for his GPUs😎

In my latest blog post, I dive deep into t-SNE: 🌐 What is t-SNE 🧠 How the algorithm works 🎯 When to use t-SNE over other dimensionality reduction techniques 📈 Advantages and Disadvantages #DataScience #MachineLearning #DataVisualization #tSNE #BigData #AI

We have a new paper (first author = Georgios Barkoulis Gavris) that leverages graph neural networks for regularized thresholding for topological optimization problems accepted by Extreme Mechanics Letter. In particular, the GNN turns a configuration with significant non-physical transition zones into a filtered configuration that allows the optimization algorithm to continue. The non-Euclidean nature of the message-passing in a graph allows the techniques to be applied to both Cartesian grid as well as polygon meshes. #topolgocialoptimization #GNN #thresholding Manuscript available: https://lnkd.in/eCKPKVrG Code and Data: https:// github.com/gsbark/TopGNN