Model training methods and devices in machine learning

Abstract

Embodiments of the present invention provide a model training method in machine learning. The method includes: extracting features from training samples; using the extracted features to train a pre-built target model, and the nonlinear part of the target model is wherein, V∈R k×P , λ∈R k×k and D ∈ R P×P is the model parameter of the nonlinear part in the target model, X∈R P×N Represents the feature matrix of the training sample, X=(x 1 ,x 2 ,...x i ,...x N ), x i ∈ R P Represents the feature vector of the i-th training sample, and x i ∈ R P is a P-dimensional column vector, λ∈R k×k and D ∈ R P×P is a diagonal matrix. In addition, another aspect of the present invention provides a model training device in machine learning.

Description

technical field [0001] Embodiments of the present invention relate to the field of machine learning, and more specifically, embodiments of the present invention relate to a model training method and device in machine learning. Background technique [0002] In the process of machine learning, it is necessary to use multiple training samples to train the training model multiple times, and finally obtain a model whose accuracy meets the predetermined requirements, that is, the ideal model. Wherein, before training, the model parameters in the training model are unknown, but after training, the model parameters in the training model are determined and known. [0003] It is understandable that during the training process, the more model parameters, the slower the training speed, and the more computing resources are consumed in training; conversely, the fewer model parameters, the faster the training speed, and the less computing resources are consumed in training. In addition, t...

AI Technical Summary

Problems solved by technology

In addition, the more model parameters, the more prone to "overfitting" problems, conversely, the fewer model parameters, the more prone to "underfitting" problems

Images