Meta-Learning Series Data Prediction with Task Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing meta-learning methods fail to achieve sufficient performance for series data, particularly when a small number of learning data is available.
Innovation Solution
A learning method that involves receiving a series data set, sampling tasks and subsets, generating task vectors using a first neural network, predicting values using a second neural network, and updating parameters to minimize errors between predicted and actual values, allowing for high-performance prediction models for series data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If existing meta-learning methods are used, then learning can be performed with a small number of learning data, but sufficient performance cannot be achieved for series data
Solution Approach 1:
The patent segments the series data into multiple tasks, where each task consists of a support set (for learning) and a query set (for evaluation). This segmentation allows the model to learn from multiple different tasks with limited data per task, while still achieving good performance through meta-learning across tasks.
Solution Approach 2:
The patent uses parameter updates through meta-learning to adapt the model to new tasks. The model parameters are optimized across multiple tasks using a small number of samples per task, enabling the system to achieve good performance with limited learning data by learning effective parameter initialization and update rules.
2Reliability
If task-specific learning data sets are used, then high performance can be achieved, but it costs a lot to prepare a sufficient amount of learning data for each task
Solution Approach 1:
The patent creates a universal meta-learning model that can be applied across multiple different tasks. Instead of training separate models for each task, a single model learns to adapt to various tasks through meta-learning, reducing the need for task-specific data preparation while maintaining high performance.
Solution Approach 2:
The patent performs preliminary meta-learning training across multiple tasks to establish good initial parameters and learning rules. This preliminary action enables the model to quickly adapt to new tasks with minimal data, avoiding the need for extensive task-specific data preparation.
3Device complexity
If a single neural network is used, then the structure is simple, but it cannot capture complex task characteristics in series data
Solution Approach 1:
The patent segments the neural network into two distinct components: a first neural network for generating task vectors from support sets, and a second neural network for predicting query values. This segmentation allows each network to specialize in its function, capturing complex task characteristics while maintaining reasonable structural simplicity.
Solution Approach 2:
The patent introduces task vectors as an intermediary between the support set and the query prediction. These task vectors capture essential characteristics of each task, enabling the model to handle complex series data patterns without requiring an overly complex network structure.
Data Source
AI summary
A learning method, executed by a computer, according to one embodiment includes an input procedure for receiving a series data set set X={Xd}d<sup2>∈</sup2>D composed of series data sets Xd for learning in a task d∈D when a task set is set as D, a sampling procedure for sampling the task d from the task set D and then sampling a first subset from a series data set Xd corresponding to the task d and a second subset from a set obtained by excluding the first subset from the series data set Xd, a generation procedure for generating a task vector representing characteristics of the first subset using parameters of a first neural network, a prediction procedure for calculating, from the task vector and series data included in the second subset, a predicted value of each value included in the series data using parameters of a second neural network, and a learning procedure for updating learning target parameters including the parameters of the first neural network and the parameters of the second neural network using an error between each value included in the series data and the predicted value corresponding to each value.