Talmaj Marinc

TL;DR: A universal compression algorithm for DNNs that is based on applying Context-based Adaptive Binary Arithmetic Coder (CABAC) to the DNN parameters, which applies a novel quantization scheme that minimizes a rate-distortion function while simultaneously taking the impact of quantization to theDNN performance into account.

TL;DR: In this article, a deep neural network based approach called Multi-Kernel Prediction Networks (MKPN) is proposed for burst image denoising, which predicts kernels of not just one size but of varying sizes and performs fusion of these kernels resulting in one kernel per pixel.

TL;DR: Using this novel method set, qualitative and quantitative analyses of the biases and artifacts in ImageNet are provided and it is demonstrated that the usage of these insights can give rise to improved models and functionally cleaned data corpora.

TL;DR: DeepCABAC is presented, a novel context-adaptive binary arithmetic coder for compressing deep neural networks that quantizes each weight parameter by minimizing a weighted rate-distortion function, which implicitly takes the impact of quantization on to the accuracy of the network into account.