Deep Learning Research

Problem 1. Literature review (3-4 pages, Times New Roman 11, single space, excluding graphics, figures and formulas): Summarize the relevant literature (2 - 3 titles what makes about 1 page per title plus references after your review) pertaining to the following topics:
1. Using LSTM for Time-Series Analysis.
2. DL Architectures for Processing Sensor Data.
3. GANs for Medical Images.
4. Embedding Autoencoders in More Complex DL Architectures and Problem Implementations.
The topic can be as narrow or as wide as you want. I expect that you would review recent research papers (published after 2016) on the topic proposing nontraditional topologies of Your report should contain citations to the respective research paper (for example [1]).
You should include: a) novelty proposed; b) if new algorithm short description; c) if new topology you may include a figure; d) information about dataset(s) used in experiments; e) comments on the results.

Problem 2. Combined deep learning architectures using RNNs or multiple GRUs as a component. Make a summary and analysis of 2-3 concrete existing architectures for solving particular problems, discussed after 2016 (required 3 to 4 pages text, excluding graphics, figures and formulas))