Neural networks are the foundation of modern deep learning. They are inspired by the human brain and can approximate complex patterns in data. Today, you'll dive into how they work, understand key concepts like activation functions and backpropagation, and get hands-on using a basic neural network implementation.
By the end of the day, you’ll have a solid grasp of neural network intuition and be ready to build and train one from scratch or using frameworks like Keras.
- What is a Neural Network?
- Perceptron and Multi-layer Perceptron
- Activation Functions (ReLU, Sigmoid, Tanh)
- Forward and Backward Propagation
- Loss Functions and Optimization
Exercise: Build a Neural Network from Scratch
- Watch the 3Blue1Brown video to understand the intuition behind neural nets.
- Follow this tutorial to implement a neural network in Python without any ML libraries.
- Alternatively, use Keras to build a simple neural net on the MNIST dataset.
- Try different activation functions and observe their impact.
- Visualize loss over epochs to see how the model learns.
Neural Networks from Scratch – Victor Zhou
Main resource for today
Neural Networks - A Visual Introduction
Interactive visual guide to neural nets by R2D3
Deep Learning with Python and Keras
Hands-on guide to neural networks using Keras
Activation Functions Explained
Comparison and usage of common activation functions
Mark today's task as complete to track your progress and earn achievements.