#ai Created at 2023-01-15 # [Anonymous feedback](https://www.admonymous.co/louis030195) # [[Epistemic status]] #shower-thought Last modified date: 2023-01-15 Commit: 0 # Related - [[Computing/Intelligence/Machine Learning/Graph neural networks]] - [[Computing/Intelligence/Joint embedding]] - [[Readwise/Articles/thegradient.pub - Transformers Are Graph Neural Networks]] # TODO > [!TODO] TODO # Unitary neural networks Unitary neural networks are a type of artificial neural network that uses unitary matrices to keep the magnitude of output activations constant. In other words, these networks use unitary matrices to ensure that the output of the network remains the same even if the input changes. This allows the network to learn from data more efficiently and retain its learned knowledge better. Unitary neural networks are used in various applications such as natural language processing, speech recognition, and computer vision. Furthermore, they can be used to improve the performance of deep learning models by increasing the speed and accuracy of training.