M
Marcin Moczulski
Researcher at University of Oxford
Publications - 19
Citations - 993
Marcin Moczulski is an academic researcher from University of Oxford. The author has contributed to research in topics: Artificial neural network & Reinforcement learning. The author has an hindex of 10, co-authored 19 publications receiving 900 citations. Previous affiliations of Marcin Moczulski include University of Michigan & Google.
Papers
More filters
Proceedings ArticleDOI
Deep Fried Convnets
Zichao Yang,Marcin Moczulski,Misha Denil,Nando de Freitas,Alexander J. Smola,Le Song,Ziyu Wang +6 more
TL;DR: A novel Adaptive Fastfood transform is introduced to reparameterize the matrix-vector multiplication of fully connected layers of deep convolutional neural networks, resulting in what is called a deep fried convnet.
Posted Content
Deep Fried Convnets
Zichao Yang,Marcin Moczulski,Misha Denil,Nando de Freitas,Alexander J. Smola,Le Song,Ziyu Wang +6 more
TL;DR: In this article, a Fastfood layer is proposed to replace all fully connected layers in a deep convolutional neural network, which substantially reduces the memory footprint of CNNs trained on MNIST and ImageNet.
Proceedings Article
Noisy activation functions
TL;DR: In this article, the authors propose to exploit the injection of appropriate noise so that the gradients may flow easily, even if the noiseless application of the activation function would yield zero gradients.
Posted Content
Noisy Activation Functions
TL;DR: This work proposes to exploit the injection of appropriate noise so that the gradients may flow easily, even if the noiseless application of the activation function would yield zero gradient, and establishes connections to simulated annealing, making it easier to optimize hard objective functions.
Proceedings Article
Contingency-Aware Exploration in Reinforcement Learning
TL;DR: This study develops an attentive dynamics model (ADM) that discovers controllable elements of the observations, which are often associated with the location of the character in Atari games, which confirms that contingency-awareness is indeed an extremely powerful concept for tackling exploration problems in reinforcement learning.