I’m currently doing B.S in Computer Science and I have to submit the proposal for my final year project in this month. I’ve been thinking a lot about this for a while now but I can’t seem to focus on or pinpoint one specific area for my project.
The only thing I’m sure of is that I want to do my project in Deep Learning. There are times when I’m more intrigued by Computer Vision problems and then there are times when I want to do something in NLP. Then there are some amazing things happening in RNN and GANs which make me quite excited.
Because of this, I am unable to choose a discipline. I’m hoping that some of you will be able to give a better advice as to how I should proceed and what are some promising areas to focus on. I’m open to any suggestions.
You made a mistake not telling what your capability is and what do you know so far. People reading the post would have to guess your interests inside the deep learning and give you their ideas. More often than not you either don’t our people’s ideas or can’t implement them.
I think Wasserstein GAN is quite interesting and a lot of problems can be solved with this kind of neural networks and no one knows when do GAN networks perform better than CNNs. So go ahead and apply GAN networks to few classic problems like object detection, text detection from images, object classification etc.
I don’t have as much experience with Deep Learning as I’d like it to be. I have completed Machine Learning Engineer Nanodegree by Udacity and also, some of the MOOCs by Andrew Ng, Machine Learning, and the new Deep Learning Specialization. I’m halfway through the part 1 of fast.ai course.
I haven’t read that many research papers, but I’ve read a lot of articles on Medium and some other sites. I’m starting to read the research papers as well.
I have a rough understanding of GANs and RNNs and just know about their successes and achievements by reading some articles as to how it’s giving some promising results at OpenAI and such.
That’s why I asked an open-ended question because I’m not so sure where to focus my efforts and I’m open to all possibilities.
Don’t go too far for a bachelor thesis. How about a full circle of crawling specific datafrom the web and then building a CNN image classifier afterwards? You have to come up with a good idea what to crawl for, but that would allow you to have a topic thats partly original without being too complex.
If you’re new to Deep Learning go with CNNs rather than RNNs, because you will need to explain precisely how they work. LSTMs are rather difficult to understand and sometimes hard to train. GANs are pretty advanced too.
I agree. I would look at doing something end to end, that can be completed in time. Also focus on making the implementation clean and having clear documentation - increase the chance of continuing to work after the course project deadline. CNN’s are a good starting point.
I would look at the Final projects from Stanford’s CS229 Machine learning course for inspiration.
Another idea would be to implement a subset of results from a recent machine learning paper, something that you are interested in. You could always extend this implementation depending on time and interest. For eg. this is a detailed relevant paper with a lot of modular sub-steps: RAISR: Rapid and Accurate Image Super Resolution
Don’t spend too much time on selecting the project to work on, treat this as a way to master skills you can use for better things ahead.
