*Understanding loss functions and the mathematical framework that drives neural network training* — ## The Foundation of Optimization The slide introduces the topic of Optimization, setting up our framework for understanding loss functions and weight matrices. But for the purpose of today, we’re mostly going to extract all those away and just think about the loss function as an abstract function. It inputs the weight matrix and outputs this scalar value of the loss. During…
Read more

*A deep dive into multiclass SVM loss with practical examples and mathematical insights* — ## The Foundation: Multiclass SVM Loss The multiclass SVM loss function ensures the score of the correct class should be higher than all other scores. This fundamental principle drives how we evaluate and optimize classification models, creating a robust framework for distinguishing between multiple categories. The loss function is visualized with a graph showing the relationship between the highest score among…
Read more

*Understanding the fundamental building blocks that power modern AI systems* — ## The Building Blocks of Neural Networks One of the most basic blocks that you’re going to have in your toolbox when you build large complicated neural networks is a linear classifier, illustrated here with stacked building blocks showing the layered structure. Much of the intuition and technical bits that we’ll cover today will carry over completely to the neural network systems that we’ll…
Read more

*Exploring how nearest neighbor algorithms work through visual analysis and the transition to k-nearest neighbors for improved classification* — ## What Does Nearest Neighbor Classification Look Like? The slide titled ‘What does this look like?’ displays results of nearest neighbor classification on the CIFAR-10 dataset through a grid of related images. Here what we’re showing is the results of nearest neighbor classification on the CIFAR 10 data set. Grid of image rows showing test images…
Read more

*Exploring the revolutionary architectures that power today’s generative AI systems* — ## The Evolution That Changed Everything Language modeling has evolved over the years. The recent breakthroughs in the past 10 years include the usage of neural networks to represent text, such as Word2Vec and N-grams in 2008. Timeline evolution In 2014, the development of sequence-to-sequence models, such as RNNs and LSTMs, helped improve the performance of ML models on NLP tasks, such as translation…
Read more