ML Model Store for IoT Deployment and User-Guided Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems lack an efficient method for training and deploying machine learning (ML) models on Internet-of-Things (IoT) devices, limiting their functionality and usability.
Innovation Solution
A ML model store system that allows for training ML models using user input and media data, and facilitates deployment on IoT devices by matching model profiles with device capabilities, enabling user selection and execution of trained models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ML models are trained using custom training data and user input, then the model accuracy and functionality for specific tasks are improved, but the complexity of the system and time required for training increase
Solution Approach 1:
The patent introduces a cloud-based ML model training system as an intermediary between the user and the model training process. The cloud system handles the complex tasks of data processing, model training, and deployment, while the edge device only needs to collect data and execute trained models. This mediator approach resolves the contradiction by transferring system complexity from the edge device to the cloud infrastructure.
Solution Approach 2:
The system segments the ML model training process into separate phases: data collection at the edge device, data transfer to cloud, model training in cloud, and model deployment back to edge. This segmentation allows each component to focus on specific tasks, reducing the complexity burden on any single device while maintaining high model accuracy through specialized cloud processing.
2Adaptability or versatility
If ML models are trained using user-provided training data, then the model becomes more adapted to specific user needs, but the time required for training and data collection increases
Solution Approach 1:
The system performs preliminary actions by pre-training base models on large datasets in the cloud before deployment to edge devices. This preliminary training creates a foundation of knowledge that can be quickly adapted to specific user needs through fine-tuning with user-provided data, significantly reducing the total training time compared to training from scratch while maintaining high adaptability.
Solution Approach 2:
The system uses parameter changes through transfer learning, where pre-trained models are fine-tuned by adjusting parameters with user-specific training data. This allows the model to adapt to specific user needs by modifying parameters rather than retraining the entire model, reducing time loss while maintaining adaptability.
3Ease of operation
If the system provides a user interface for model selection and training, then the ease of operation is improved, but the device complexity increases
Solution Approach 1:
The cloud-based training system serves as an intermediary that handles the complex operations of model management, training, and deployment through web interfaces. This allows users to interact with sophisticated ML capabilities through simple graphical interfaces without needing to understand the underlying system complexity, effectively resolving the contradiction between ease of operation and system complexity.
4Measurement precision
If the system collects and processes media data for model training, then the quality of training data is improved, but the energy consumption and data processing requirements increase
Solution Approach 1:
The cloud infrastructure acts as an intermediary that handles the energy-intensive data processing and model training operations. The edge device only performs lightweight data collection and transmission tasks, while the cloud handles the computationally intensive processing. This distribution resolves the contradiction by moving energy consumption to the cloud where it can be managed more efficiently.
Solution Approach 2:
The system extracts the data processing and training operations from the edge device and places them in the cloud infrastructure. By taking out the energy-intensive tasks from the edge device, the system maintains high training data quality through sophisticated cloud processing while significantly reducing the energy consumption requirements at the edge device level.
Data Source
AI summary
In one aspect, an example method involves: receiving a request to train a model and prompting a user for a first input indicating a subject for detection within media; receiving the first input; using at least the received first input as a basis to obtain a set of media related to the subject for detection; outputting the obtained set of media and prompting the user for second input indicating subject identification information; receiving the second input; using at least (i) the obtained set of media as training input data and (ii) the received second input as training output data, to train the model; and performing operations to facilitate causing a computing system to run the trained model, wherein the computing system running the trained model comprises the computing system using at least the trained model and received runtime input data to generate and output corresponding runtime output data.