17847080. DATA STORAGE DEVICE WITH NOISE INJECTION simplified abstract (Western Digital Technologies, Inc.)

From WikiPatents
Jump to navigation Jump to search

DATA STORAGE DEVICE WITH NOISE INJECTION

Organization Name

Western Digital Technologies, Inc.

Inventor(s)

Daniel Joseph Linnen of Naperville IL (US)

Kirubakaran Periyannan of Saratoga CA (US)

Ramanathan Muthiah of Bangalore (IN)

Grant Chapman Mackey of Laguna Hills CA (US)

DATA STORAGE DEVICE WITH NOISE INJECTION - A simplified explanation of the abstract

This abstract first appeared for US patent application 17847080 titled 'DATA STORAGE DEVICE WITH NOISE INJECTION

Simplified Explanation

The patent application describes procedures for injecting noise into a non-volatile memory (NVM) array. This is done to induce bit flips and information degradation in the data stored in the memory. The injected noise can be used for dataset augmentation or for testing deep neural networks (DNNs).

  • Noise is injected into data stored in a non-volatile memory array by adjusting read voltages to induce bit flips.
  • Feedback is used to achieve a target amount of information degradation.
  • Random data is combined with itself iteratively to achieve a target percentage of random 1s or 0s.
  • The random data is then combined with data read from the NVM array.
  • Pixels in the array can be randomly zeroed out to simulate dead charge coupled device (CCD) pixels.
  • Timing, voltage, and/or current values used in circuits during data transfer are adjusted outside their specified margins to induce bit flips and inject noise into the data.

Potential Applications

  • Dataset augmentation for machine learning and deep neural networks.
  • Testing and evaluation of deep neural networks.
  • Simulation of faulty or degraded memory conditions for testing purposes.

Problems Solved

  • Lack of diverse and realistic datasets for training machine learning models.
  • Difficulty in testing and evaluating the robustness of deep neural networks.
  • Limited ability to simulate faulty or degraded memory conditions for testing purposes.

Benefits

  • Improved performance and accuracy of machine learning models through dataset augmentation.
  • Better understanding of the robustness and reliability of deep neural networks.
  • More realistic testing of memory systems and algorithms.


Original Abstract Submitted

Noise injection procedures implemented on the die of a non-volatile memory (NVM) array are disclosed. In one example, noise is injected into data by adjusting read voltages to induce bit flips while using feedback to achieve a target amount of information degradation. In another example, random data is iteratively combined with itself to achieve a target percentage of random 1s or 0s, then the random data is combined with data read from the NVM array. In other examples, pixels are randomly zeroed out to emulate dead charge coupled device (CCD) pixels. In still other examples, the timing, voltage, and/or current values used within circuits while transferring data to/from latches or bitlines are adjusted outside their specified margins to induce bit flips to inject noise into the data. The noise-injected data may be used, for example, for dataset augmentation or for the testing of deep neural networks (DNNs).