Locality, Learning, and the FFT: Why CNNs Avoid the Fourier Domain
A unified treatment of 1D signal convolution, 2D image convolution via the convolution theorem, and graph convolution as a spectral operation on the normalized Laplacian.
Read more
Part III : What does Low Rank Factorization of a Convolutional Layer really do?
In this post, we will explore the Low Rank Approximation (LoRA) technique for shrinking neural networks for embedded systems. We will focus on the Convolutional Neural Network (CNN) case and discuss the rank selection process.
Read morePart II : Shrinking Neural Networks for Embedded Systems Using Low Rank Approximations (LoRA)
In this post, we will explore the Low Rank Approximation (LoRA) technique for shrinking neural networks for embedded systems. We will focus on the Convolutional Neural Network (CNN) case and discuss the rank selection process.
Read moreRecents
