Hyperparameter Tuning with PyTorch for Neural Networks & Computer Vision: the bias variance tradeoff

An introduction to conceptual tools you'll need to produce the best possible machine learning model you can using hyperparameter tuning with PyTorch. Takes a step by step approach to doing model tuning with a basic neural network for a computer vision task.

Background videos:

Bounds:    • Видео  
Bias & Variance:    • Видео  
Neural Network Architectures:    • Видео  

A complete free online ML course: https://intelligentsystem.io/course.html

Data set: http://parnec.nuaa.edu.cn/xtan/data/d...

Demonstrates two important hyperparameter search methods, including what you'll see during the process, and how to interpret and adapt. Concepts include: bounds, bias variance tradeoff, generalization, neural networks, over-fitting and under-fitting, parameter sweeps, grid search, directed search, computer vision, convolutional networks, cross validation, data augmentation, normalization, convergence properties and more.

The approach covered in this video is a basis for doing professional applied machine learning, going from a data set to a well-tuned and effective model. The general concepts apply to any modeling approach (not just to neural networks).

This is part of my Lost Lectures series, which covers topics I wish I'd had time to get to when I taught a graduate level machine learning course.