Multi-Task Model Input Gating to Avoid Redundant Predictions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning models waste computer resources by executing unnecessary tasks on data, leading to redundant prediction results.
Innovation Solution
A method where first information is added to the input of a machine learning model to determine specific tasks needed on the data, allowing adaptive execution and reducing redundant predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a machine learning model executes N tasks simultaneously on input data, then the model can provide comprehensive multi-task processing capability, but redundant prediction results are generated when not all N tasks are needed, causing waste of computer resources
Solution Approach 1:
The patent applies dynamics by making the task execution configuration adjustable and adaptive. The system dynamically determines which tasks to execute based on input data characteristics and user requirements, rather than rigidly executing all N tasks. This allows the model to adapt its processing scope to match actual needs, reducing redundant computations while maintaining multi-task capability when required.
Solution Approach 2:
The patent implements local quality by applying different processing approaches to different tasks. Instead of uniformly executing all tasks on all data, the system selectively applies task processing based on local conditions - specifically, whether a particular task is needed for a given input data item. This localized task execution eliminates redundant predictions while preserving necessary multi-task processing.
2Reliability
If a machine learning model generates N prediction results for N tasks, then complete task coverage is achieved, but unnecessary prediction results are produced when only M tasks (M<N) are actually needed
Solution Approach 1:
The patent applies partial action by executing only the necessary subset of tasks (M tasks) rather than all N tasks. The system determines the appropriate scope of task execution based on actual requirements, performing just enough processing to meet needs without over-processing. This partial execution maintains reliability for required tasks while improving productivity by avoiding unnecessary computations.
Solution Approach 2:
The patent implements preliminary action by determining task requirements before executing the machine learning model. The system预先 (in advance) identifies which tasks are needed based on input data characteristics and user preferences, then configures the model to execute only those tasks. This preliminary determination prevents redundant execution and optimizes the balance between completeness and efficiency.
3Stability of the object's composition
If all N tasks are executed on every input data item, then consistent task coverage is maintained, but resource consumption increases unnecessarily when some tasks are not needed
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the task execution configuration parameters based on input data characteristics and user requirements. Instead of maintaining a fixed configuration that executes all N tasks, the system modifies the active task parameters to match actual needs. This parameter adaptation maintains consistent task coverage for required tasks while reducing energy consumption by deactivating unnecessary tasks.
4Adaptability or versatility
If a machine learning model is designed to handle multiple tasks, then flexibility in task selection is improved, but complexity of task configuration and management increases
Solution Approach 1:
The patent applies universality by designing a unified machine learning model that can handle multiple tasks through a common processing framework. Instead of creating separate configurations for each task combination, the system uses a universal model structure with selectable task modules. This universal approach provides flexible task selection while managing complexity through standardized interfaces and unified task management mechanisms.
Data Source
AI summary
A data processing method, a model training method, and a related device are provided. The method may be applied to a multi-task processing scenario in the field of artificial intelligence. The method includes: obtaining first data and first information, where the first information indicates at least one task executed on the first data; and inputting the first data and the first information into a first machine learning model, and processing the first data by using the first machine learning model, to obtain at least one prediction result that is output by the first machine learning model and that is in a one-to-one correspondence with the at least one task. To be precise, the first machine learning model can learn of, based on the first information, specific tasks that need to be executed on the first data, so that a required task can be adaptively executed on the first data.