Deep-Learning

Today, I am going to start my grand tour in deep learning. Review what I have learnt and try new things.

Datasets:

Image:

1. coco
2. flower
3. bird
4. CIFAR-10 and 100

NLP:

Model:

CNN

1. VGG
2. PixelCNN
3. Yolo

RNN (LSTM),attention

GAN:

1. DCGAN
2. AttnGan
3. CycleGAN

Transformers:

1. Implement from scratch [Done]
2. Other transformers

Diffusion models

1. Stable diffusion

Tasks:

1. Classification
2. Object detection a) YOLO b) R-CNN c) SSD
3. Segmentation a) Masked R-CNN b) DETR c) Segment Anything

Deep RL: value based, policy gradient, actor critic, ...

1. DQN [Done]

Read papers:

Research directions:

visual and language intergration (express them in numbers with same sequence of number expressing same meaning)

Device a common network to process those information (self decide how many parameter to include in the model)

Small points

1. Optimization methods
2. Stableize methods