Automatic Attribute Extraction for ML Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning model training methods are resource-intensive and time-consuming, often requiring large amounts of labeled data and expensive processing resources like GPUs, which can lead to inefficiencies and lower accuracy in model training.
Innovation Solution
The proposed solution involves using semi-supervised based approaches combined with iterative methods such as stochastic gradient descent (SGD) to train machine learning models, allowing for faster convergence of cross-entropy losses and more efficient use of processing resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning with manually created labels is used, then training accuracy is improved, but processing resources and time consumption increase
Solution Approach 1:
The patent introduces an intermediary attribute extraction system that automatically generates labels from unlabelled data using a trained machine learning model. This intermediary layer translates unlabelled data into labelled data without requiring manual annotation, thereby reducing the time and resource investment while maintaining training accuracy through the use of automatically extracted attributes
2Measurement precision
If more training data is collected to improve model accuracy, then model performance is improved, but processing resources and training time increase
Solution Approach 1:
The system enables self-service label generation where the machine learning model itself extracts attributes and generates labels from unlabelled data. This self-service mechanism allows the system to transform unlabelled data into useful training labels automatically, reducing the need for manual intervention and decreasing the processing resources required to handle large volumes of training data
3Productivity
If expensive processing resources like GPUs are employed to reduce training time, then training speed is improved, but cost increases
Solution Approach 1:
The patent segments the training process into two distinct phases: a training phase where the machine learning model is trained on labelled data, and an inference phase where the trained model extracts attributes from unlabelled data. This segmentation allows the system to use less expensive processing resources during the inference phase, reducing overall hardware costs while maintaining efficient training speeds through the structured approach
Data Source
AI summary
This application relates to apparatus and methods for training machine learning models using supervised, or semi-supervised, learning. In some examples, a computing device obtains training data that includes labelled, and unlabeled, data for training a machine learning model. The computing device applies the machine learning model to the training data to generate output data. The machine learning model executes with a plurality of coefficients applied to a plurality of hyperparameters. The computing device further applies a loss model to the training data and the output data to generate a loss value. Based on the loss values, the computing device determines updated values for the plurality of coefficients of the machine learning model. The computing device may continue to determine updated values for the plurality of coefficients until one or more conditions are satisfied. The computing device may then store the final coefficient values in a data repository.