Fact Finder
Top 20
Behzad Benam
Sep 16
·
2 min read
Optimizing System Resources by Using Neuromorphic Computing
Neuromorphic systems will replace GPUs in the future
Spiking Neural Networks (SNN) are artificial neural networks based on biological knowledge to optimize the resources required to run machine learning algorithms. They are very similar to biological neural networks and the mechanism of brain operation.
The human brain has about 100 billion neurons and hundreds of trillions of connections. The fastest GPUs with the same network size require about half a gigawatt power. Therefore, we need to reduce the power consumption of machine learning algorithm execution, which is one of the barriers to realizing humanoid robot technology. Neuromorphic computing takes the human brain as a reference to optimize resources for machine learning algorithms.
Energy efficiency
Neuromorphic technology is more power efficient than GPU-based artificial neural networks. Energy consumption is a crucial issue for large networks. Neuromorphic computing aims to mimic our brain's low-power computing ability. Due to the low power consumption of neuromorphic hardware, measuring power consumption is challenging and requires a new strategy.
Traditional computer architecture, known as von Neumann architecture, physically separated the system into memory units, central processing units, and other units. This separation causes power inefficiency and is a significant limitation for the future of computing systems. The human brain behaves differently because we know that memory and control units in our brain are not separate, and both are together.
Self-organizing and self-learning
Neuromorphic computing can solve uncertainties by learning new subjects like the human brain. Recently presented approaches in robotics can create new algorithms. The neuromorphic architecture is not as dependent on high-quality training data as deep neural networks.
Execution performance
According to Moore's Law, the number of transistors in a microchip doubles every two years. Although Moore's Law is not entirely correct and the speed of doubling transistors has changed, based on new research, neuromorphic computing is a way to keep Moore's Law still true.
View attachment 20952
WOW.
Peter van der Made may have been speaking the truth at the AGM in 2019 when he addressed shareholders.
WOW.
Maybe this is why an electrical engineer was not prepared to do his own research???
Wolves in sheep clothing???
My opinion only DYOR as @Bravo does many thanks.
FF
AKIDA BALLISTA
