Название: Hyperparameter Tuning for Machine and Deep Learning With R: A Practical GuideАвтор: Eva Bartz, Thomas Bartz-Beielstein. Martin Zaeffere
Издательство: Springer
Год: 2023
Страниц: 327
Язык: английский
Формат: pdf (true), epub
Размер: 43.3 MB
Hyperparameter tuning? Is this relevant in practice? Is it not rather an academic gimmick? This book provides a wealth of hands-on examples that illustrate how hyperparameter tuning can be applied in practice and gives deep insights into the working mechanisms of Machine Learning (ML) and Deep Learning (DL) methods. Programming code is provided so that users can reproduce the results. The aim of the book is to equip readers with the ability to achieve better results with significantly less time, costs, effort and resources using the methods described here. The case studies presented in this book can be run on a regular desktop or notebook computer. No high-performance computing facilities are required.
ML and DL methods are becoming more and more important and are used in many industrial production processes, e.g., Cyber-physical Production Systems (CPPS). Several hyperparameters of the methods used have to be set appropriately. Previous projects carried out produced inconsistent results in this regard. For example, with Support Vector Machines (SVMs) it could be observed that the tuning of the hyperparameters is critical to success with the same data material, with random forests the results do not differ too much from one another despite different selected hyperparameter values. While some methods have only one or a few hyperparameters, others provide a large number. In the latter case, optimization using a (more or less) fine grid (grid search) quickly becomes very time-consuming and can therefore no longer be implemented. In addition, the question of how the optimality of a selection can be measured in a statistically valid way (test problem: training/validation/test data and resampling methods) arises for both many and a few hyperparameters. In real-world projects, DL experts have gained profound knowledge over time as to what reasonable hyperparameters are, i.e., Hyper Parameter Tuning (HPT) skills are developed. These skills are based on human expert and domain knowledge and not on valid formal rules.
Similar to the example in Chap. 10, which considered tuning a Deep Neural Network (DNN), this chapter also deals with neural networks, but focuses on a different type of learning task: reinforcement learning. This increases the complexity, since any evaluation of the learning algorithm also involves the simulation of the respective environment. The learning algorithm is not just tuned with a static data set, but rather with dynamic feedback from the environment, in which an agent operates. The agent is controlled via the DNN. Also, the parameters of the reinforcement learning algorithm have to be considered in addition to the network parameters. Based on a simple example from the Keras documentation, we tune a DNN used for reinforcement learning of the inverse pendulum environment toy example. As a bonus, this chapter shows how the demonstrated tuning tools can be used to interface with and tune a learning algorithm that is implemented in Python.
As in Chap. 10, we use Keras and TensorFlow to implement the neural networks. However, we will perform the complete learning task within Python, using the R package reticulate to explicitly interface between the R-based tuner and the Python-based learning task (rather than implicitly via R ’s keras package). On the one hand, this will demonstrate how to interface with different programming languages (i.e., if your model is not trained in R). On the other hand, this is a necessary step, because the respective environment is only available in Python (i.e., the toy problem).
Скачать Hyperparameter Tuning for Machine and Deep Learning With R: A Practical Guide