Marcel Marais’ Post

Great work is often not recognized or appreciated in the moment. Such was the case of the support vector machine (SVM), the crown jewel of classical machine learning. Here is the story ... Way back in 1964, in the then U.S.S.R, Vladimir Vapnik and Alexey Chervonenkis invented a classification method based on maximum margin hyperplanes. That was a whopping four decades before the era of fast GPU processors which ushered in the deep learning era. In 1990, Vapnik immigrated to the United States to work at AT&T Bell labs. There, he collaborated with Bernhard Bohser and Isabella Guyon, to apply the "kernel trick" to his 1964 maximum margin hyperplane method, thereby yielding the powerful support vector machine we know today. Excited about this great new discovery, in 1992 Vapnik submitted 3 papers describing the SVM to NeurIPS, the premier machine learning conference. All 3 papers were very swiftly rejected by the reviewers. Today, the SVM stands as a towering monument in the world of classical machine learning, bridging powerful ideas, frameworks, and eras. It turned out that the reviewers were squarely wrong. Great work is often not recognized or appreciated in the moment. If you are working on something today and are not getting the recognition and support you believe your work deserves, think of Vapnik and keep going. You are in good company. ---- Stay tuned for the entire video series of all 20 chapters from my book "The Foundational Mathematics of Artificial Intelligence." See comments for details. #ArtificialIntelligence #Mathematics

This is astonishing. GPT 4 "discovered" this sorting algorithm without any training. Some of you would know how important it is to find a better/faster sorting algorithm. It is at the foundation of most computing. And to think that the world's computer scientists, engineers, and mathematicians could not come up with this one, which is up to 70% faster than algorithms in current use. The last sorting algorithm was created in 2006 (library sort). https://lnkd.in/drC3aHq6

Just finished Build GANs and Diffusion Models with TensorFlow and PyTorch! Check it out: https://lnkd.in/djxY7N7m #machinelearning #artificialintelligence #generativeadversarialnetworks

If you're looking for the best introductory course to dive into Machine Learning, I have an incredible YouTube playlist: This series breaks down ML concepts in an approachable way and covers a wide range of topics, including: 📚 Linear Regression: The foundation of predictive modeling. 📊 Classification: From logistic regression to advanced classifiers. 📈 Decision Trees: Easy-to-visualize yet powerful models. ⚙️ Model Evaluation: Understanding metrics like accuracy, precision, and recall. 🔢 Clustering and Dimensionality Reduction: Techniques for unsupervised learning and simplifying data. What makes this playlist stand out is its clarity and step-by-step approach, making it ideal for beginners. It’s a great starting point whether you're a student, professional, or just curious about ML! I highly recommend bookmarking this playlist and working through it at your own pace. Let me know if you've checked it out or have other recommendations to share! #MachineLearning #AI #DataScience #LearningResources

I was so lucky to learn support vector machines from the ground up. It is indeed the hallmark of classical machine learning and foundational to so many of the models we see and know today.

SVM's are definitely one of the more interesting supervised learning models. Maximizing the margin while minimizing classification error using hinge loss.

Just finished Build GANs and Diffusion Models with TensorFlow and PyTorch! Check it out: https://lnkd.in/gqRK8uCw #machinelearning #artificialintelligence #generativeadversarialnetworks

Just completed: Build GANs and Diffusion Models with TensorFlow and PyTorch! https://lnkd.in/ecZ-jwcY #machinelearning #artificialintelligence #generativeadversarialnetworks”