18218893. METHODS AND ELECTRONIC DEVICE FOR REPAIRING MEMORY ELEMENT IN MEMORY DEVICE simplified abstract (Samsung Electronics Co., Ltd.)

METHODS AND ELECTRONIC DEVICE FOR REPAIRING MEMORY ELEMENT IN MEMORY DEVICE

Organization Name

Samsung Electronics Co., Ltd.

Inventor(s)

Helik Kanti Thacker of Bengaluru (IN)

Adrita Barari of Bengaluru (IN)

Akhilesh Sudhir Patankar of Bengaluru (IN)

Deokgu Yoon of Suwon-si (KR)

Damini Damini of Bengaluru (IN)

Keerthi Kiran Jagannathachar of Bengaluru (IN)

Paulami Das of Bengaluru (IN)

Sairam Jujjarapu of Bengaluru (IN)

Sudhanshu Gupta of Bengaluru (IN)

METHODS AND ELECTRONIC DEVICE FOR REPAIRING MEMORY ELEMENT IN MEMORY DEVICE - A simplified explanation of the abstract

This abstract first appeared for US patent application 18218893 titled 'METHODS AND ELECTRONIC DEVICE FOR REPAIRING MEMORY ELEMENT IN MEMORY DEVICE

Simplified Explanation

The abstract of this patent application describes a method for repairing a memory element in a memory device using an electronic device. The method involves representing the memory element as a graph with a vertex and an edge, where the node associated with the memory element contains information about a fault. A repair policy is determined from the graph using a probability distribution over faulty and non-faulty lines, predicted by a graph neural network (GNN) based on the final node feature value obtained from message passing stages of the GNN. The value of a state is determined based on the probability of the memory element being repaired from a particular state, using a global mean of all the final node feature values predicted by the GNN.

• The memory element is represented as a graph with a vertex and an edge.
• The node associated with the memory element encodes information about a fault.
• A repair policy is determined using a probability distribution over faulty and non-faulty lines.
• The repair policy is predicted by a graph neural network (GNN) based on the final node feature value obtained from message passing stages of the GNN.
• The value of a state is determined based on the probability of the memory element being repaired from a particular state.
• The probability is calculated using a global mean of all the final node feature values predicted by the GNN.

Potential applications of this technology:

• Memory device repair in electronic devices.
• Fault detection and repair in computer systems.

Problems solved by this technology:

• Efficient repair of memory elements in memory devices.
• Improved fault detection and repair in electronic devices.

Benefits of this technology:

• Enhanced reliability and performance of memory devices.
• Reduction in downtime and maintenance costs for electronic devices.

Original Abstract Submitted

A method for repairing a memory element in a memory device by an electronic device includes configuring a memory element as a graph with a vertex and an edge, a node associated with the memory element being encoded with information related to a fault, determining, from the graph, a repair policy using a probability distribution over one or more of a faulty line and a non-faulty line as predicted by a graph neural network (GNN) based on a final node feature value from message passing stages of the GNN, and determining a value of a state using a probability of the memory element being repaired from a particular state based on a global mean of all the final node feature values predicted by the GNN.