Model training method, model predication method, model training device, model predication method, electronic device and machine learning platform

Abstract

The invention provides a model training method, model predication method, model training device, model predication method, electronic device and machine learning platform. The model training method includes loading a description file of an algorithm, wherein the description file of the algorithm includes description information used for describing dependency relation among a plurality of algorithmprocessing steps; loading an execution file of the algorithm according to the description file of the algorithm and performing model training according to the training data, wherein the execution file of the algorithm includes execution files of different algorithm processing steps in the algorithm. The invention enables customization of a model training process for a user.

Description

technical field [0001] The present invention relates to the field of computers, in particular to a model training and prediction method, device, electronic equipment and machine learning platform. Background technique [0002] Machine learning (Machine Learning, ML) is a multi-field interdisciplinary subject, specializing in the study of how computers simulate or implement human learning behaviors to acquire new knowledge or skills, and reorganize existing knowledge structures to continuously improve their own performance. . [0003] Machine learning usually includes the following steps 101-104: [0004] 101. Obtain the training data set, and preprocess the data, such as normalizing, removing singular points, etc.; [0005] 102. Select machine learning methods, strategies (that is, how to define the loss function) and algorithms (Algorithm); among them, in the field of machine learning, algorithms refer to the theoretical logic of solving a certain type of specific problem...

AI Technical Summary

Problems solved by technology

[0009] Usually, only the built-in model of the platform can be used for training and prediction, and users cannot develop their own algorithms, and cannot support customized model training and prediction, and cannot meet the needs of various business scenarios; when updating the algorithm, it is necessary to stop training, so it is difficult to do to the dynamic update of the algorithm; and the model trained offline cannot be directly used for online prediction, and the user needs to load the trained model into the prediction system by himself during prediction

Images