Distributed ML Model Selection for Real-Time Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large deep neural networks consume significant memory and processing resources, take time to process datasets, and require large training datasets, making them unsuitable for real-time operations.
Innovation Solution
A distributed data processing system utilizing a computing device and additional devices to execute multiple smaller machine learning models, each trained for specific input datasets, with a load balancer and mode selector to determine the most efficient model for processing based on features and accuracy comparisons.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large deep neural network is used to achieve high accuracy, then the accuracy is improved, but the memory consumption and processing resources increase significantly
Solution Approach 1:
The patent divides a large deep neural network into multiple smaller sub-networks, each trained on specific subsets of data. These smaller networks process different segments of the input data in parallel, achieving high accuracy while reducing individual model memory consumption and enabling distributed processing across multiple devices.
Solution Approach 2:
The patent transitions from a single large model architecture to a distributed multi-model architecture, adding the dimension of spatial distribution across multiple computing devices. This allows the system to maintain high accuracy through ensemble processing while reducing the resource burden on any single device.
2Measurement precision
If a large deep neural network is used to achieve high accuracy, then the accuracy is improved, but the processing time increases
Solution Approach 1:
By segmenting the large network into smaller sub-networks that process different data segments in parallel, the system reduces the sequential processing time while maintaining overall accuracy through the aggregation of results from multiple smaller models.
Solution Approach 2:
The patent merges the results from multiple smaller sub-networks to achieve the final prediction. This combining approach allows parallel processing of data segments, significantly reducing total processing time while maintaining the accuracy benefits of a large network through ensemble decision-making.
3Measurement precision
If a large deep neural network is used to achieve high accuracy, then the accuracy is improved, but the training time increases
Solution Approach 1:
The training process is segmented into multiple independent training tasks, where each smaller sub-network is trained separately on specific data subsets. This parallel training approach significantly reduces total training time compared to training one large network sequentially, while the ensemble of trained models maintains high accuracy.
4Use of energy by moving object
If multiple smaller machine learning models are used to reduce resource consumption, then the resource efficiency is improved, but the system complexity increases
Solution Approach 1:
The patent introduces a load balancer as an intermediary component that manages the complexity of coordinating multiple smaller models. The load balancer handles model selection, data routing, and result aggregation, abstracting the system complexity while enabling efficient resource utilization across distributed computing devices.
5Speed
If multiple computing devices are used to process data in parallel, then the processing speed is improved, but the coordination complexity increases
Solution Approach 1:
The load balancer serves multiple functions simultaneously: it acts as a model selector, data router, result aggregator, and resource manager across distributed computing devices. This multi-functional intermediary simplifies coordination complexity while enabling parallel processing to achieve high processing speeds.
Data Source
AI summary
A distributed data processing system and a distributed data processing method are provided. The distributed data processing system includes a computing device and at least one additional computing device. In the method, the computing device receives a dataset to be processed, selects at least one computing device capable of executing a first machine learning model on the dataset from the computing device and the additional computing device to process the dataset and generate a first prediction result, selects at least one computing device capable of executing each of multiple machine learning models on the dataset from the computing device and the additional computing device to process the dataset and generate at least a second prediction result and a third prediction result, selects a second machine learning model from the multiple machine learning models based on a comparison between the first prediction result and at least the second prediction result and the third prediction result, and instructs the at least one computing device for executing the second machine learning model to process the dataset.