Taking the Human out of Learning Applications: A Survey on Automated Machine Learning
Definition of AutoML?
The current classification of AutoML?
What are the methods?
What technologies are available now
What is the status quo
Q: Why do you need AutoML?
A:
It is mainly in the process of machine learning, without human interference, directly completing the steps from data input processing to prediction output.
AutoML is the intersection of machine learning and automation.
There are two points of view:
1. Regarding AutoML itself as a powerful machine learning tool
. 2. Regarding AutoML as building a high-level control system to complete all work without human intervention.
Q:AutoML的定义
A:
max configurations performance of learning tools s . t . { no human assistance limited computational budget \max_{\text{configurations}} \text{performance of learning tools} \\ s.t.\left\{ \begin{array}{l} \text{no human assistance} \\ \text{limited computational budget} \end{array} \right. configurationsmaxperformance of learning toolss.t.{
no human assistancelimited computational budget
The general model of AutoML to solve the problem
Classification of current AutoML work