Welcome to our Articles Page!

These articles mostly contain summaries of interesting Deep Learning papers, as well as opinion posts on certain topics. Please Check it out!

Medical Image Augmentation

How Generative Adversarial Networks can improve Medical Image datasets

Class-Splitting GAN

Explaining how class splitting can improve the results of Generative Adversarial Networks

StyleGAN

Explaining how Adaptive Instance Normalization is used in Generative Adversarial Networks to achieve incredible control over the latent space

Searching for Activation Functions

Using AutoML to design Activation Functions

Maxout Networks

Dicussing an Activation Function designed by Ian Goodfellow in 2013

Are GANs Created Equal?

Experiment showing that hyperparameter optimization results in equal performance across different loss functions

DenseNets

CNN Connectivity Pattern that achieves state-of-the-art results on Image Classification

Cutout Regularization

Randomly cutting out patches of images to regularize CNNs and reduce overfitting

DeepWalk

Using ideas of Word2Vec to construct Graph embeddings and make predictions based on connectivity

Inception Network

CNN Connectivity Pattern that increases the width of the network rather than the depth

StackGAN

Text-to-Image synthesis using a multi-scale architecture and augmentation to the state-of-the-art text and visual embedding algorithm

SR DenseNets

Applying the DenseNet connectivity pattern to Super-Resolution

HyperNetworks

Using Meta-Learning to predict the initialization of Deep Neural Networks. This has huge potential for other Meta-learning methods such as NAS