User-customizable machine learning model with sensor data and label input
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Consumers with limited technical knowledge lack the ability to customize and utilize machine learning models to suit their specific needs or circumstances, as they typically only have access to trained results without the means to tailor them according to their unique requirements.
Innovation Solution
A user-customizable machine learning model system that includes sensors to monitor user, object, and environment aspects, allowing users to input user-defined labels via a user interface, forming a training set, and training a machine learning model to predict applicable labels for further measurements, enabling personalized insights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning models are made accessible to consumers, then the ability to customize and utilize ML models for specific needs is improved, but the complexity of the system increases
Solution Approach 1:
The system segments the machine learning workflow into distinct components: sensor data collection, user label input through interface, training set formation, and model training. This segmentation allows consumers to interact with only the necessary parts (data input and label definition) while the complex processing occurs automatically in the background, resolving the contradiction between accessibility and complexity.
Solution Approach 2:
A processing unit acts as an intermediary between the consumer and the machine learning system. The processing unit automatically handles complex tasks including forming training measurement sets from sensor data, training the ML model, and generating predictions. The user interface serves as another intermediary layer that simplifies user interaction, allowing consumers to define labels without understanding the underlying complexity.
2Measurement precision
If users are provided with tools to customize ML models, then the precision and relevance of predictions for specific user needs is improved, but the ease of operation decreases
Solution Approach 1:
The system enables self-service by allowing users to define their own labels and categories based on their specific needs. Users can input custom labels through the user interface, and the processing unit automatically handles the entire training process. This self-service capability improves prediction precision for user-specific requirements while maintaining ease of operation through automated processing.
Solution Approach 2:
The system performs preliminary actions by automatically forming training measurement sets from sensor data before the user needs to make predictions. The processing unit pre-processes the data, creates the training set, and trains the model in advance, so that when users need predictions, the system is already prepared. This eliminates the need for users to perform complex data preparation tasks.
3Adaptability or versatility
If the system collects and processes user data for custom training, then the relevance of predictions to user circumstances is improved, but the loss of time for data collection and processing increases
Solution Approach 1:
The system implements continuous data collection from sensors in the background while the user interacts with the application. Sensor measurements are continuously gathered and stored, and the processing unit can train the ML model at any time using the accumulated data. This continuous operation eliminates idle time and ensures that training can occur immediately when sufficient data is available, reducing the overall time loss.
Data Source
AI summary
An apparatus for providing a user-customisable machine learning model, MLM, the apparatus including: (i) one or more processing units configured to: receive measurements from one or more sensors that monitor one or more aspects of a user, an object, or an environment of the user; (ii) receive, via a user interface, one or more user-defined labels for one or more parts of the received measurements; (iii) form a training measurement set from the one or more parts and the one or more user-defined labels; and (iv) train an MLM with the training measurement set to generate a trained MLM, wherein the trained MLM is such that the trained MLM is able to determine one or more user-defined labels for further measurements received from the one or more sensors.