Multi-Encoder Neural Network for Simultaneous Domain Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning technologies face challenges in achieving high precision and low costs when simultaneously learning multiple tasks and domains, as they often neglect feature expressions unique to each task or domain, leading to limited applicability and reduced precision.
Innovation Solution
A model learning device and method that utilizes a combination of target, source, and common encoders and decoders to learn feature expressions specific to different tasks and domains, allowing for the simultaneous learning of multiple tasks and domains by matching output results with training data, using a multilayer-structure neural network configuration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a model learns using data from a single domain, then precision for that domain is improved, but adaptability to other domains deteriorates
Solution Approach 1:
The patent segments the feature expression into two distinct components: domain-specific feature expressions (unique to each domain) and task-specific feature expressions (unique to each task). This segmentation allows the model to separately learn and combine domain-specific and task-specific features, resolving the contradiction between precision for a specific domain and adaptability to multiple domains.
Solution Approach 2:
The patent creates a universal model structure that can handle multiple domains and tasks simultaneously by incorporating both domain-specific and task-specific feature expressions. The model becomes multi-functional, capable of performing different tasks across different domains while maintaining high precision through the specialized feature expressions for each domain-task combination.
2Adaptability or versatility
If feature expressions common to multiple domains are used, then adaptability is improved, but precision for specific tasks deteriorates
Solution Approach 1:
The patent segments feature expressions into domain-specific and task-specific components, preventing the model from relying solely on generic common features. By separating these components, the model can maintain adaptability through the shared model structure while achieving high precision through domain- and task-specific feature expressions.
Solution Approach 2:
The patent applies local quality by making different parts of the model have different characteristics: the model structure provides universal adaptability, while the domain-specific and task-specific feature expressions provide localized precision for each domain and task combination.
3Measurement precision
If separate models are created for each task, then precision for each task is improved, but device complexity and cost increase
Solution Approach 1:
The patent merges multiple separate models into a single unified model that handles multiple tasks and domains simultaneously. By combining domain-specific and task-specific feature expressions within one model framework, the system achieves the precision of multiple specialized models while reducing the complexity and cost of maintaining and training separate models for each task.
Solution Approach 2:
The patent creates a universal model structure that can perform multiple tasks across different domains, eliminating the need for separate specialized models. This multi-functional approach reduces device complexity and training costs while maintaining high precision through the integrated domain-specific and task-specific feature expressions.
4Measurement precision
If domain-specific models are trained, then precision for that domain is improved, but time and computational cost for training increase
Solution Approach 1:
The patent merges the training of multiple domain-specific models into a single unified training process. By training domain-specific and task-specific feature expressions simultaneously within one model framework, the system reduces total training time and computational cost while maintaining the precision benefits of domain-specific training.
Solution Approach 2:
The patent performs preliminary action by pre-training domain-specific feature expressions that can be reused across multiple tasks. This allows the model to leverage previously learned domain knowledge when training for specific tasks, reducing the overall training time and computational resources required.
Data Source
AI summary
Simultaneous learning of a plurality of different tasks and domains, with low costs and high precision, is enabled. A learning unit 160, on the basis of learning data, uses a target encoder that takes data of a target domain as input and outputs a target feature expression, a source encoder that takes data of a source domain as input and outputs a source feature expression, a common encoder that takes data of the target domain or the source domain as input and outputs a common feature expression, a target decoder that takes output of the target encoder and the common encoder as input and outputs a result of executing a task with regard to data of the target domain, and a source decoder that takes output of the source encoder and the common encoder as input and outputs a result of executing a task with regard to data of the source domain, to learn so that the output of the target decoder matches training data, and the output of the source decoder matches training data.