← Search

Alex Sergeev

2 accepted papers

2018

An intriguing failing of convolutional neural networks and the CoordConv solution

NeurIPS 2018poster

Few ideas have enjoyed as large an impact on deep learning as convolution. For any problem involving pixels or spatial representations, common intuition holds that convolutional neural networks may be appropriate. In this paper we show a striking counterexample to this intuition via the seemingly tr…

2018

Faster Neural Networks Straight from JPEG

NeurIPS 2018poster

The simple, elegant approach of training convolutional neural networks (CNNs) directly from RGB pixels has enjoyed overwhelming empirical success. But can more performance be squeezed out of networks by using different input representations? In this paper we propose and explore a simple ide…