• Publications
  • Influence
Understanding and Improving Convolutional Neural Networks via Concatenated Rectified Linear Units
This paper proposes a novel, simple yet effective activation scheme called concatenated ReLU (CRelu) and theoretically analyze its reconstruction property in CNNs and integrates CRelu into several state-of-the-art CNN architectures and demonstrates improvement in their recognition performance on CIFAR-10/100 and ImageNet datasets with fewer trainable parameters. Expand
Improved Multimodal Deep Learning with Variation of Information
This paper proposes a novel multimodal representation learning framework that explicitly aims to minimize the variation of information, and applies this framework to restricted Boltzmann machines and introduces learning methods based on contrastive divergence and multi-prediction training. Expand
ELF: An Extensive, Lightweight and Flexible Research Platform for Real-time Strategy Games
ELF, an Extensive, Lightweight and Flexible platform for fundamental reinforcement learning research, is proposed and it is shown that a network with Leaky ReLU and Batch Normalization coupled with long-horizon training and progressive curriculum beats the rule-based built-in AI more than $70\% of the time in the full game of Mini-RTS. Expand
Exploring Normalization in Deep Residual Networks with Concatenated Rectified Linear Units
A simple modification to NormProp is proposed and the modified NormProp performs better than the original NormProp but is still not comparable to BatchNorm, and CReLU improves the performance of ResNets with or without normalizations. Expand
Convolutional Neural Networks for Crop Yield Prediction using Satellite Images
Crop yield forecasting during the growing season is useful for farming planning and management practices as well as for planning humanitarian aid in developing countries. Common approaches to yieldExpand
Channel-Recurrent Autoencoding for Image Modeling
This work proposes two novel regularizations, namely the KL objective weighting scheme over time steps and mutual information maximization between transformed latent variables and the outputs, to enhance the training of channelrecurrent VAE-GAN. Expand
Learning World Graphs to Accelerate Hierarchical Reinforcement Learning
A thorough ablation study is performed to evaluate the proposed graph abstraction over the environment structure to accelerate the learning of these tasks with significant advantages from the proposed framework over baselines that lack world graph knowledge in terms of performance and efficiency. Expand
Channel-Recurrent Variational Autoencoders
This paper proposes to integrate recurrent connections across channels to both inference and generation steps of VAE, and shows that the channel-recurrent VAE improves existing approaches in multiple aspects. Expand
Advantage Actor-Critic Methods for CarRacing
Autonomous driving, a complex integration of perception, planning and control, is one of today’s most prominent, fast-growing technology. Recent works have proposed to pose the learning of autonomousExpand