It is reported that the Korean Institute of Science and Technology (KIST) recently announced that a joint research group led by Dr. Song Jingmi, Dr. Zhu Xianzhang, Director Zhang Junyan and Dr. Wu Shengxun of IBM Corporation has developed the world's first use at its next-generation semiconductor research institute. The core component of skyrmion's neuromorphic computer. Researchers predict that if the part is made smaller and several of them are interconnected, it will drive the development of artificial intelligence (AI) CPUs.
The team used this artificial synaptic material to conduct an improved National Institute of Standards and Technology (MNIST) recognition test and found that it achieved a high recognition rate of 90% through only 15,000 learnings, while other artificial synaptic materials required Hundreds of thousands of iterations can reach 90% recognition rate. This means that the technology developed by the team requires less than 10% of the power.
Dr. Song explained, "This composition is very similar to the human brain. This component controls the weight of synapses through the number of electrically controlled skyrmions, thereby regulating synapses through the number of neurotransmitters."
It is understood that the results of this study were published in the online edition of the global journal Nature Electronics on March 16.
The team used this artificial synaptic material to conduct an improved National Institute of Standards and Technology (MNIST) recognition test and found that it achieved a high recognition rate of 90% through only 15,000 learnings, while other artificial synaptic materials required Hundreds of thousands of iterations can reach 90% recognition rate. This means that the technology developed by the team requires less than 10% of the power.
Dr. Song explained, "This composition is very similar to the human brain. This component controls the weight of synapses through the number of electrically controlled skyrmions, thereby regulating synapses through the number of neurotransmitters."
It is understood that the results of this study were published in the online edition of the global journal Nature Electronics on March 16.