Neural Network Knowledge Module for Data Protection

The patent extracts knowledge from training data records by training neural network models that learn patterns and representations. The trained models encapsulate the essential knowledge without containing the original personal data, allowing knowledge retention while removing data protection risks. This is achieved through the training process where models learn from data and then the original data can be deleted while preserving the learned knowledge in the model parameters.

The patent creates abstract representations or copies of the knowledge contained in training data through neural network models. These models serve as simplified copies that capture the essential patterns and relationships without replicating the original personal data. The models can be shared and reused across different applications, enabling knowledge transfer while maintaining data protection since the models themselves do not contain personal information.

The patent performs preliminary knowledge extraction by training neural network models before deleting the original training data. The training process captures and stores knowledge in the model parameters and weights. Once the models are trained and the knowledge is embedded in them, the original training data can be safely deleted while the knowledge remains preserved in the model structures, enabling compliance with data protection regulations.

The patent transforms the training data into a different parameter representation through the neural network training process. The original data parameters are converted into model parameters (weights, biases, architecture configurations) that represent the same knowledge in an abstract form. This parameter transformation allows the knowledge to be retained in a format that does not reveal personal information, enabling data deletion while preserving knowledge.