AI Materials

---

CV

---

• CNN

  • The pioneering work of CNN. The paper introduces the convolutional layer, pooling layer, fully connected layer, and the implementation of the CNN model.

• transformer

  • The pioneering work of the transformer. The paper introduces the self-attention mechanism and the multi-head attention mechanism, as well as the implementation of the transformer model.

• GAN

  • The pioneering work of GAN, which introduces the generator and discriminator mechanism in GAN.

• cGAN

  • Reference Blog

    • Help build a network from scrat

  • Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks

  • ganhacks

  • pix2pix

• Unsupervised GAN

  • Refernce Blog

    • A detailed explanation of GAN and its implementation, as well as the unsupervised GAN.

  • Tutorial: Generative Adversarial Networks

  • CycleGAN (style transfer)

  • CycleGAN (Chinese font transfer)

---

RNN

---

• word2vec

  • A detailed explanation of word2vec and its implementation CBOW & SG, as well as two optimization methods, negative sampling and hierarchical softmax.

---

RL

• DQN

  • The pioneering work of DQN, which introduces the experience replay mechanism and the target network mechanism.

• A3C

  • The pioneering work of A3C, which introduces the actor-critic mechanism and the asynchronous training mechanism.

• Converge

  • Prove the convergence and stability of the algorithm.

• DDPG

---

Tips

---

• CUDA restart

sudo rmmod nvidia_uvm
sudo modprobe nvidia_uvm