[this page on wiki][index][EN][BG][CS][DA][DE][EL][ES][ET][FI][FR][GA][HR][HU][IT][MT][NL][PL][PT][RO][SK][SL][SV]

Lecture: Hyperparameter tuning

Administrative Information

Title Hyperparameter tuning
Duration 60
Module B
Lesson Type Lecture
Focus Technical - Deep Learning
Topic Hyperparameter tuning

Keywords

Hyperparameter tuning, activation functions, loss, epochs, batch size,

Learning Goals

Expected Preparation

Obligatory for Students

None.

Optional for Students

None.

References and background for students

  • John D Kelleher and Brain McNamee. (2018), Fundamentals of Machine Learning for Predictive Data Analytics, MIT Press.
  • Michael Nielsen. (2015), Neural Networks and Deep Learning, 1. Determination press, San Francisco CA USA.
  • Charu C. Aggarwal. (2018), Neural Networks and Deep Learning, 1. Springer
  • Antonio Gulli,Sujit Pal. Deep Learning with Keras, Packt, [ISBN: 9781787128422].

Recommended for Teachers

None.

Lesson materials

Instructions for Teachers

This lecture will introduce students to the fundamentals of the hyperparameter tuning. We will use the Census Dataset as the examples of the use and outcomes from tuning various hypermeters. The Adult Census dataset is a binary classification problem. More on this dataset in the corresponding tutorial. The goal of this lecture is to introduce several hyperparameters with examples of how modifying these hyperparameters may aid or hinder learning. In addition we provide examples of under and overfitting, nose and performance gains (training time and in some cases accuracy/loss) when each of the hyperparameters are tunned. We will use diagnostic plots to evaluate the effect of the hyperparameter tunning and in particular a focus on loss, where it should be noted that the module we use to plot the loss is matplotlib.pyplot, thus the axis are scaled. This can mean that significant differences may appear not significant or vice versa when comparing the loss of the training or test data. In addition some liberties for scaffolding are presented, such as the use of Epochs first (almost as a regularization technique) while keeping the Batch size constant. Ideally these would be tunned together, but for this lecture they are separated.

Outline

Hpt.png
Time schedule
Duration (Min) Description
5 Overview of the data
10 Capacity and depth tunning (under and over fitting)
10 Epochs (under and over training)
10 Batch sizes (for noise suppression)
10 Activation functions (and their effects on performance - time and accuracy)
10 Learning rates (vanilla, LR Decay, Momentum, Adaptive)
5 Recap on the forward pass process

Acknowledgements

The Human-Centered AI Masters programme was Co-Financed by the Connecting Europe Facility of the European Union Under Grant №CEF-TC-2020-1 Digital Skills 2020-EU-IA-0068.