Udemybeginneren

Machine Learning: Neural networks from scratch

Learn to build neural networks from the ground up using Python! Master core concepts and implement them yourself. Perfect for deep learning enthusiasts.

4.1(21)
129students
4h 54m
Certificate
Maxime Vandegar
Maxime VandegarIngénieur de recherche

Course Content

Machine Learning: Neural networks from scratch is a beginner-level Udemy course taught by Maxime Vandegar. It runs 4h 54m, is taught in English, and includes a certificate of completion. Learners rate it 4.1 out of 5 from 21 reviews, with 129 students enrolled. It is a paid course, last updated October 2024.

Dive into the fascinating world of neural networks with this comprehensive course! Learn to build a neural network from scratch, gaining a deep understanding of its inner workings without relying on external libraries. While we'll use Python for implementation, the core concepts are transferable, empowering you to implement neural networks in any programming language.

This course takes you on a journey from intuitive understanding to rigorous mathematical foundations. You'll explore essential techniques for stabilizing neural network training, including the crucial log-sum-exp trick and the Jacobian-vector product method to manage memory usage. These techniques are vital for successfully training modern neural networks.

Get hands-on experience by training your custom-built neural networks on real-world image classification and regression problems. We'll implement a variety of cost functions (like MSELoss and NLLLoss) and explore different activation functions (including ReLU, Softmax, and LogSoftmax). You'll also learn to create reusable modules for building complex neural network architectures.

Key concepts covered include:

  • Neural Networks Fundamentals: A thorough exploration of the core principles.
  • Implementing Neural Networks from Scratch: Build your own network!
  • Gradient Descent & Jacobian Matrix: Master optimization techniques.
  • Modular Design: Create nested modules for complex architectures.
  • Log-Sum-Exp Trick: Stabilize training for better results.
  • Jacobian-Vector Product: Manage memory and prevent exploding gradients.
  • Activation Functions: ReLU, Softmax, LogSoftmax, and more.
  • Cost Functions: MSELoss, NLLLoss, and other essential loss functions.

This course is perfect for developers eager to understand and implement neural networks at a fundamental level, and for anyone seeking a comprehensive "A to Z" understanding of how these powerful models function.

Prerequisites: Basic Python programming skills are required. If you're new to programming, consider a crash course. A foundational understanding of Algebra and Calculus will greatly enhance your learning experience.

Stay tuned for frequent updates and bonus content! Start your machine learning journey today!

$ whoami — the devs behind tutorialsearch.io

Want AI in your product? Hire the team behind TutorialSearch.io

LLM features, RAG, agents and automation wired into a real product — built by the devs who shipped this site.