Why would AutoML automated machine learning?
1. Machine learning feature extraction, data analysis project is also characterized by the most time-consuming part of the work effort, unlike algorithms and models are determined steps, and in most cases have to rely on personal experience algorithm engineers to deal with. This feature works with strong uncertainties, such as leakage characteristics of the election, the election is invalid to the characteristics, and other advanced features ignored.
2. In order to solve the above problems born AutoML, AutoML trying to project these features, model selection and parameter optimization and other important steps to automate learning, making the machine learning model can be applied without human intervention.
Visible from the front section, the generalization of machine learning has been restricted by many conditions will need for a more general solution to solve the problem, which resulted AutoML. AutoML is a will fundamentally change the status quo solutions based on machine learning programs.
AutoML is a neural network control can raise a child model architecture training and performance evaluation, the test results will feedback on specific tasks to the controller to let the controller know how the next round to improve their models. Automatic machine learning focused on two areas: data collection and model predictions. In between these two phases all the steps that occur will be learning machine automatically abstracted. In fact, users only need to provide their own data set, label and press a button to generate the prediction and optimization model of a fully trained. Most platforms prompt the user to upload the data set, then category labeled. Thereafter, most of the steps involved in the data preprocessing, model selection, and super engineering parameters optimization features will be processed in the background. This approach has greatly changed the traditional workflow involved in training machine learning models.
AutoML completely changed the rules of the game the whole field of machine learning, because for many applications, does not require specialized skills and knowledge. Many companies need to complete only the depth of the network simpler tasks such as image classification. So they do not need to hire some of the artificial intelligence expert, they just need to get better data organization, and then handed over to AutoML to be completed.
Hard vegetables, slowly digestion;
