Apple developers recently proposed a machine learning algorithm called DeepPCR, which can be used to accelerate neural network inference and training efficiency.
This machine learning algorithm is designed to accelerate the processing efficiency of neural network calculations, avoiding the process of scheduling neural networks to execute tasks sequentially, which may increase the overall training and feedback generation time due to the accumulation of computing data. By reducing the complexity of the cyberlink process through a parallel cycle reduction (PCR) calculation method, the overall calculation time is shortened.
After Apple internally deployed and used the DeepPCR algorithm, the forward transfer speed of training could be increased by up to 30 times, and the backward transfer speed could be increased by up to 200 times, thereby accelerating the efficiency of neural network inference and training.
Apple has not yet publicly announced its progress in artificial intelligence, emphasizing that it is not lagging behind in applications such as machine learning. However, there have been numerous reports that Apple is actively investing in AI technology, with the goal of enhancing the user experience of applications and services such as Siri.
