Neural Acquisition Training for Fast Unknown Function Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing active learning methods face challenges such as high training costs, difficulty in selecting acquisition criteria, and optimizing acquisition functions, especially when learning time is constrained, making it difficult to efficiently determine a model for an unknown function.
Innovation Solution
A method involving training a neural network to select inputs for evaluating an unknown function through multiple iterations of sampling Gaussian processes, using the neural network to evaluate initial guesses, and adjusting its parameters to improve an objective function, allowing for efficient active learning without extensive real data usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional active learning methods are used to train models for unknown functions, then the model can be determined through iterative sampling and evaluation, but the training process becomes computationally expensive and time-consuming
Solution Approach 1:
The patent creates a synthetic training environment by sampling Gaussian processes to generate virtual data points and function evaluations. This synthetic copy allows the neural network to learn acquisition function optimization without requiring actual costly function evaluations, thereby reducing training time while maintaining model determination accuracy.
Solution Approach 2:
The patent performs preliminary training of the neural network using synthetic data generated from sampled Gaussian processes. This preliminary action prepares the acquisition function before it needs to be applied to real unknown functions, allowing fast inference without retraining when actual function evaluations are needed.
2Reliability
If multiple acquisition criteria are evaluated to select the best one, then the active learning process becomes more robust, but the complexity of selecting and optimizing acquisition functions increases
Solution Approach 1:
The patent trains the neural network to automatically select and optimize acquisition functions through self-supervised learning on synthetic data. The network learns to evaluate and compare different acquisition criteria implicitly during training, eliminating the need for manual selection and complex optimization procedures while maintaining robust performance.
Solution Approach 2:
The patent replaces manual acquisition function optimization with a neural network that learns the optimization process through synthetic training. This substitution eliminates complex mechanical optimization procedures and manual selection processes, reducing computational complexity while improving reliability through automated learning.
3Measurement precision
If real data collection is performed for training the model, then the model reflects actual system behavior, but the cost and time required for data collection increases
Solution Approach 1:
The patent generates synthetic training data by sampling Gaussian processes, creating virtual copies of actual system behavior without requiring real data collection. These synthetic data points maintain the statistical properties and functional relationships of real systems while eliminating the cost and time of actual data collection, though they approximate rather than exactly replicate real system behavior.
Data Source
AI summary
A method for determining a model for an unknown function is described comprising training a neural network for selecting inputs at which to evaluate the unknown function. The training includes a plurality of iterations of sampling, from a set of Gaussian processes, at least one initial guess for the unknown function, using the neural network to select inputs and evaluating the selected inputs using the at least one initial guess, determining a value of an objective function from the evaluated selected inputs, adjusting the neural network to improve the value of the objective function and determining the model by evaluating the unknown function at a sequence of inputs given by the trained neural network and fitting the model to the evaluated inputs.