Anfonwch hwn fel neges destun: Speeding up convolutional neural networks with low rank expansions