17942064. PRODUCT AUTOENCODER FOR ERROR-CORRECTING VIA SUB-STAGE PROCESSING simplified abstract (Samsung Electronics Co., Ltd.)

From WikiPatents
Jump to navigation Jump to search

PRODUCT AUTOENCODER FOR ERROR-CORRECTING VIA SUB-STAGE PROCESSING

Organization Name

Samsung Electronics Co., Ltd.

Inventor(s)

Mohammad Vahid Jamali of Ann Arbor MI (US)

Hamid Saber of San Diego CA (US)

Homayoon Hatami of San Diego CA (US)

Jung Hyun Bae of San Diego CA (US)

PRODUCT AUTOENCODER FOR ERROR-CORRECTING VIA SUB-STAGE PROCESSING - A simplified explanation of the abstract

This abstract first appeared for US patent application 17942064 titled 'PRODUCT AUTOENCODER FOR ERROR-CORRECTING VIA SUB-STAGE PROCESSING

Simplified Explanation

The patent application describes a processing circuit that includes an encoder and a decoder. The encoder takes original data and encodes it using an error correction code (ECC), while the decoder takes received data and decodes it using the same ECC.

  • The encoder supplies k symbols of original data to a neural product encoder.
  • The neural product encoder consists of M neural encoder stages.
  • Each neural encoder stage is configured with specific parameters to implement an (n,k) error correction code.
  • The encoder outputs n symbols representing the k symbols of original data encoded by the ECC.
  • The decoder supplies n symbols of a received message to a neural product decoder.
  • The neural product decoder is grouped into l pipeline stages, with each pipeline stage consisting of M neural decoder stages.
  • Each neural decoder stage is configured with specific parameters to implement an (n,k) ECC.
  • The decoder outputs k symbols decoded from the n symbols of the received message.

Potential Applications:

  • Data transmission and communication systems that require error correction.
  • Storage systems that utilize error correction to ensure data integrity.
  • Any application where reliable and accurate data encoding and decoding is crucial.

Problems Solved:

  • Error correction codes are essential for data transmission and storage, but traditional methods may be computationally expensive.
  • This technology aims to provide a more efficient and effective way to implement error correction codes using neural networks.

Benefits:

  • Improved efficiency in encoding and decoding data using neural networks.
  • Reduced computational complexity compared to traditional error correction methods.
  • Enhanced reliability and accuracy in data transmission and storage systems.


Original Abstract Submitted

A processing circuit implements: an encoder configured to: supply k symbols of original data to a neural product encoder including M neural encoder stages, a j-th neural encoder stage including a j-th neural network configured by j-th parameters to implement an (n,k) error correction code (ECC), where nis a factor of n and kis a factor of k; and output n symbols representing the k symbols of original data encoded by an error correcting code; or a decoder configured to supply n symbols of a received message to a neural product decoder including neural decoder stages grouped into a l pipeline stages, an i-th pipeline stage of the neural product decoder including M neural decoder stages, a j-th neural decoder stage comprising a j-th neural network configured by j-th parameters to implement an (n,k) ECC; and output k symbols decoded from the n symbols of the received message.

Retrieved from "http://wikipatents.org/index.php?title=17942064._PRODUCT_AUTOENCODER_FOR_ERROR-CORRECTING_VIA_SUB-STAGE_PROCESSING_simplified_abstract_(Samsung_Electronics_Co.,_Ltd.)&oldid=21280"