Arithmetic Processing Unit Eigen Solution Range Prediction

The patent applies preliminary action by using a learning model to predict the eigen solution range before the actual search process. The learning model is trained on historical data of input values and their corresponding eigen solutions, allowing the system to pre-estimate the search range and only then perform the eigen solution search within this narrowed range, thereby reducing computational complexity while maintaining accuracy.

The patent changes the parameter of search range from the entire possible range to a predicted range based on learning model output. By transforming the search space parameter using the learning model's prediction, the system efficiently narrows down the eigen solution search range, reducing the number of calculations required while maintaining sufficient accuracy for the application.

The learning model performs preliminary analysis of historical input-value and eigen-solution data to predict the likely range of eigen solutions for new inputs. This preliminary action allows the system to avoid searching the entire range and instead focus computational resources on the predicted range, reducing overall computational complexity while maintaining search completeness within the relevant domain.

The learning model acts as an intermediary between the input values and the eigen solution search process. It processes historical data to create a predictive mapping that mediates the relationship between inputs and expected eigen solution ranges, enabling the system to skip irrelevant search spaces and reducing computational complexity while maintaining comprehensive search within the predicted range.

The patent uses copying by creating a learning model that replicates the patterns from historical input-value and eigen-solution data. The model learns from past examples and creates a simplified representation (copy) of the relationship between inputs and eigen solutions, which can then be used to quickly predict search ranges for new inputs without re-analyzing the entire historical dataset, thereby improving processing speed.

The learning model performs preliminary learning from historical data during an initial training phase, after which it can quickly make predictions for new inputs. This preliminary action of training the model once allows for rapid subsequent predictions, improving processing speed while the model itself becomes an integrated part of the system architecture.