NVM Die Data Augmentation in Storage Controllers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning systems require large amounts of training data and existing data augmentation methods transfer significant data volumes, which can be inefficient and costly, especially in near-memory computing architectures.
Innovation Solution
Implementing data augmentation within non-volatile memory (NVM) dies using under-the-array or next-to-the-array components, reducing the need to transfer large augmented data sets by performing data augmentation operations directly on the NVM die.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data augmentation is performed using traditional external systems, then machine learning training data variety is improved, but data transfer volume and cost increase significantly
Solution Approach 1:
The patent moves data augmentation from external systems to within the NVM die itself, utilizing the third dimension of on-chip integration. By embedding data augmentation components directly in the NVM die, the system eliminates the need to transfer large volumes of augmented data between separate systems, while still achieving diverse training data through in-situ augmentation operations.
Solution Approach 2:
The patent introduces an intermediary data augmentation component that resides between the NVM array and the external system. This intermediary performs augmentation operations on the NVM die, acting as a mediator that generates diverse training data locally without requiring extensive data transfer to external augmentation systems.
2Adaptability or versatility
If data augmentation operations are performed externally, then data variety is improved, but processing efficiency and cost decrease
Solution Approach 1:
The patent merges data augmentation functionality with the NVM die structure by integrating data augmentation components directly into the die. This consolidation allows augmentation operations to be performed alongside storage operations, eliminating separate processing steps and improving overall processing efficiency while maintaining data variety.
Solution Approach 2:
The NVM die performs data augmentation operations on its own stored data without requiring external processing systems. The data augmentation components within the die enable self-service augmentation, where the storage device itself generates diverse training data, improving processing efficiency by eliminating external transfer and processing steps.
3Quantity of substance
If large augmented data sets are transferred externally, then training data volume is improved, but system complexity and cost increase
Solution Approach 1:
The patent extracts the data augmentation function from external complex systems and places it directly within the NVM die. By taking out the augmentation operation from external processors and embedding it in the storage device, the system reduces overall system complexity while still generating sufficient training data volume through in-situ augmentation.
Data Source
AI summary
Methods and apparatus are disclosed for implementing data augmentation within a storage controller of a data storage device based on machine learning data read from a non-volatile memory (NVM) array of a memory die. Some particular aspects relate to configuring the storage controller to generate augmented versions of training images for use in training a Deep Learning Accelerator of an image recognition system by rotating, translating, skewing, cropping, etc., a set of initial training images obtained from a host device and stored in the NVM array. Other aspects relate to controlling components of the memory die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of the NVM array to inject noise into the images. Data augmentation based on data read from multiple memory dies is also described, such as image data spread across multiple NVM arrays or multiple memory dies.