Deep Learning in Audio for Absolute Beginners: From No Experience to a Deployed Model

Discover how to start with deep learning in audio processing, even as a complete beginner, in this comprehensive 51-minute conference talk from the Audio Developer Conference. Learn to debunk common myths about deep learning, acquire essential skills through a four-step process, create datasets for modeling analog devices, train your first deep learning model using PyTorch without extensive computational resources, and deploy your model in a real-time audio plugin. Follow along with a live demonstration of setting up a deep learning pipeline and training a neural network for Virtual Analog modeling of a distortion effect. Gain insights into machine learning concepts, neural network architectures, and practical techniques for audio applications, all presented in an accessible manner for those new to the field.

Introduction
Outline
My Mission
Music Maker Jam
Video Editor Jam
My Background
What is Deep Learning
Why Deep Learning
Examples
Myths
Deep Learning is easy
You need a lot of data
You need a computational cluster
Deep learning models can run in real time
Deep learning techniques
Deep learning book
Deep learning lectures
Deep learning research papers
What are the 4 steps
Out of the way
Machine Learning
Datasets
Virtual Analog Modeling
Neural Networks
Fir Filter
Recurrent Neural Network
Loss Function in Audio
Gradient Descent
HyperParameters
Epochs
Free Datasets
Demonstration
Spice Software
Code Base
Resources
Conclusion
Questions
Simple RNN
RNN Output