I have been wanting to implement a Deep Learning Paper to get some hands on the current state of the art model or current field of research. But, generally the paper I pickup is a bit tough to understand. So, I was looking if anyone could suggest me a paper which would be some latest research but slightly easier to grasp?
Ask HN: What deep learning papers should I implement to learn?
1–5 of 5 posts
Re: Ask HN: What deep learning papers should I implement to learn?
#2Pushing the barriers or independent validation on the latest research is not a good place to start if you're new to the area.
Go and implement an image classifier, tweak the parameters and the topology and see how the results and training changes.
Switch to a time series model / text model, and learn the difference between convolution / recurrent networks.
Start playing with non-sequential topologies, custom objective functions, Q-Learning,
Once you have a grasp of these basics go back and read the papers, and you'll find that you understand them a lot more, and you'll see where they're pushing the boundaries.
Re: Ask HN: What deep learning papers should I implement to learn?
#3Pushing the barriers or independent validation on the latest research is not a good place to start if you're new to the area. Go and implement an image classifier, tweak the parameters and the topology and see how the results and training changes. Switch to a time series model / text model, and learn the difference between convolution / recurrent networks. Start playing with non-sequential topologies, custom objectiv…
If he wants to focus on Deep Learning I wouldn't touch Reinforcement Learning (you mentioned Q-learning). Lots of hours can be wasted on RL.
Re: Ask HN: What deep learning papers should I implement to learn?
#4That's actually a pretty good idea to improve your skills. I would highly recommend to implement one of the semantic segmentation papers. Perhaps start with a simple FCN (with and without a pretrained encoder).
Re: Ask HN: What deep learning papers should I implement to learn?
#5Pushing the barriers or independent validation on the latest research is not a good place to start if you're new to the area. Go and implement an image classifier, tweak the parameters and the topology and see how the results and training changes. Switch to a time series model / text model, and learn the difference between convolution / recurrent networks. Start playing with non-sequential topologies, custom objectiv…
If he wants to focus on Deep Learning I wouldn't touch Reinforcement Learning (you mentioned Q-learning). Lots of hours can be wasted on RL.
Agreed!