Deep Learning Data Measurement for Accurate User-Specific Results
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data measurement methods are inefficient and lack a systematic approach to meet user requirements for data calculation, particularly in the context of big data environments.
Innovation Solution
A data measurement method utilizing a pre-trained deep learning network that includes identity verification, training of initial models, and aggregation of models to generate a combined training model, ensuring accurate and user-specific measurement results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual data calculation methods are used, then users can perform data measurement, but the efficiency is low and errors occur
Solution Approach 1:
The patent replaces manual mechanical calculation methods with an automated deep learning-based measurement system. The system uses trained neural network models to automatically process data, eliminating manual intervention and its associated errors while significantly improving measurement efficiency and consistency.
Solution Approach 2:
The system enables self-service data measurement through automated model training and execution. Users can independently train measurement models using their own data sets, and the system automatically performs measurements without requiring manual calculation processes, thereby improving both efficiency and reliability.
2Measurement precision
If a pre-trained deep learning network is used, then measurement accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent applies preliminary action by providing pre-trained deep learning models that have already been trained on extensive data sets. Users can directly utilize these pre-trained models without needing to perform complex training procedures themselves, thereby achieving high measurement accuracy while keeping the system relatively simple to deploy.
Solution Approach 2:
The system manages complexity by allowing parameter adjustments in the training process rather than changing the fundamental system architecture. Users can modify training parameters such as learning rates, batch sizes, and data set compositions to optimize measurements without redesigning the entire deep learning framework.
3Measurement precision
If model training is performed with identity verification, then user-specific accuracy is improved, but the operation time increases
Solution Approach 1:
The system performs preliminary identity verification and user authentication before initiating model training. This preliminary action ensures that the correct user-specific models are trained and applied, improving measurement precision while the verification process itself is designed to be quick and efficient.
Solution Approach 2:
The patent uses model copying by creating user-specific copies of base measurement models. Instead of training entirely new models for each user, the system starts with pre-trained base models and fine-tunes them with user-specific data, significantly reducing training time while maintaining high user-specific accuracy.
4Reliability
If multiple models are aggregated to form a combined training model, then measurement reliability is improved, but the device complexity increases
Solution Approach 1:
The patent merges multiple trained models into a single combined training model through aggregation. This combining process integrates the strengths of individual models to improve measurement reliability and robustness. The aggregation is performed systematically using defined combination strategies that manage complexity while achieving enhanced performance.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A data measurement method and apparatus, and an electronic device and a computer-readable medium. The method comprises: acquiring a data set (102); inputting the data set (102) into a pre-trained deep learning network, and outputting a processing result (103), wherein the deep learning network is obtained by training a training sample set, and training of the deep learning network comprises: in response to receiving a training request of a target user, acquiring identity information of the target user, verifying the identity information, and determining whether the verification is passed, and in response to it being determined that the identity information passes the verification, controlling a target training engine to start training; and determining the processing result (103) to be a measurement result (104), and controlling a target device having a display function to display the measurement result (104). The data set (102) is input into the pre-trained deep learning network, such that the measurement result (104) that meets user requirements can be obtained. The requirements of a user for data calculation are met, and convenience is provided for the user to subsequently use data.