Inference Model Generation via Glue Layer and Partial Model Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating inference models are often burdened with a heavy load, especially when training data sets are limited, leading to redundancy and increased computation requirements.
Innovation Solution
The proposed method generates a third inference model by combining trained portions of two inference models, specifically a first partial inference model and a second partial inference model, with a glue layer connecting intermediate layers, reducing the overall load and redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a second inference model is trained using output data from a first inference model as input data, then the inference model can be generated, but the load applied during generation becomes heavy
Solution Approach 1:
The patent divides the second inference model into multiple portions (first portion, second portion, third portion) and trains them separately using different training data sets. This segmentation reduces the computational load on any single training process while still achieving the goal of generating a functional second inference model that uses output from the first inference model as input.
2Adaptability or versatility
If trained inference models are combined to generate a new inference model, then model generation can proceed with limited training data, but redundancy increases
Solution Approach 1:
The patent extracts and removes redundant components from the combined inference model. Specifically, it identifies and eliminates duplicate or unnecessary model portions while retaining the essential functional components needed for inference, thereby reducing overall model redundancy while maintaining generation capability.
Solution Approach 2:
The patent merges multiple trained inference model portions into a unified second inference model. By combining the first portion, second portion, and third portion with complementary functions, it creates a versatile model that can handle various inference tasks using limited training data without excessive redundancy.
3Reliability
If full inference models are trained and combined, then comprehensive inference capability is achieved, but computation requirements increase
Solution Approach 1:
The patent applies partial training actions by training only specific portions of the second inference model rather than the entire model at once. The first portion is trained with first training data, the second portion with second training data, and the third portion with third training data, achieving comprehensive inference capability through multiple partial training actions rather than one excessive full-model training.
Data Source
AI summary
An inference model generating method is a method for generating a third inference model using a trained first inference model and a trained second inference model, when a type of output data that is output from the first inference model is the same as a type of input data that is input to the second inference model, the method including: preparing a first partial inference model that includes a portion of the first inference model from an input layer through a predetermined intermediate layer; preparing a second partial inference model that includes a portion of the second inference model from a predetermined intermediate layer to an output layer; and generating the third inference model by disposing a glue layer between the first partial inference model and the second partial inference model.