Скачать или смотреть Session 1 — From Scratch to TensorFlow: Building Your First Neural Network

In this video, I walk through two approaches to training a neural network on the Iris dataset — the same dataset I assign for homework. First, I manually code a simple neural network and train it for 50 epochs to show how the forward pass, loss, and weight updates work step by step. Then I switch to a TensorFlow implementation to automate the process and compare results. Along the way, I check accuracy for both training and testing sets, explore overfitting, and plot loss and accuracy curves for training and validation. Finally, I visualize the decision boundary to show how the model separates flower classes.

This tutorial is designed for teens and beginners who want to understand deep learning both from scratch and with modern tools.