andrewjong/Deep-Learning-Paper-Surveys

Todo List

Opened this issue · 0 comments

andrewjong commented

■ Deep Learning Papers

◎ General Deep Learning

  • Activation Functions

    • Swish [TO READ]

  • Batch Size

  • Learning Rate

◎ Vision

  • Hand-Designed Architectures

    • Accuracy

      • ?

    • Mobile Efficiency

      • MobileNetV1: Depthwise Separable Convolutions
      • MobileNetV2: Inverted and Linear Residual Bottlenecks
      • Fully Learnable Group Convolution for Acceleration of Deep Neural Networks
      • GhostNet

    • Squeeze-and-Excitation Networks (seems to be commonly used in a lot of top architectures)

  • Neural Architecture Search

    • MNasNet [TO WRITE]
    • Searching for MobileNetV3 [TO READ]
    • EfficientNets [TO WRITE]
    • Designing Network Design Spaces [TO WRITE]
    • FBNet [TO WRITE]
    • FBNetV2

  • Data Augmentation

    • AutoAugment [TO READ]
    • Fast AutoAugment [TO READ]
    • AdvProp [TO READ]

  • Camera Pipeline

    • Deep Demosaicing for Edge Implementation

◎ Generative Graphics

  • Networks

    • DCGAN [TO WRITE]
    • StyleGAN [TO WRITE]
    • MSG-GAN [TO READ]
    • StyleGAN2 [TO READ]

  • Layers

    • Spectral Normalization GAN (SN-GAN) [TO READ]

  • Loss Functions

    • WGAN [TO WRITE]
    • WGAN-GP [TO WRITE]
    • Mescheder R1/R2 [TO WRITE]
    • Implicit Competitive Regularization [TO READ]

  • Systems (fundamental systems that generalize widely)

    • Pix2Pix [TO WRITE]
    • CycleGAN [TO WRITE]
    • Guided Image-to-Image Translation with Bi-Directional Feature Transformation [TO READ]

  • Applications

    • Virtual Try-on
      • FW-GAN [TO WRITE]

◎ Reinforcement Learning

◎ Natural Language Processing

  • Architectures
    • Attention is All You Need (Transformers) [TO READ]
    • BERT [TO READ]