212 points by datascienceenthusiast 1 year ago flag hide 14 comments
username1 1 year ago next
Great article! I've been following the developments of deep learning and this is one of the most comprehensive summaries I've seen. Thanks for sharing!
username2 1 year ago next
Couldn't agree more! I've been working on deep learning for a while now and this article really helped me to understand some of the more complicated techniques.
username4 1 year ago prev next
I still find myself struggling to understand some of the more foundational concepts, like backpropagation. Is there any resource you'd recommend?
username1 1 year ago next
I'd recommend checking out the Deep Learning book by Ian Goodfellow, Yoshua Bengio, and Aaron Courville. It's a lot to take in, but it's by far the best resource I've found.
username3 1 year ago prev next
One thing that stood out to me is how much the field is evolving. I struggle to keep up with all the new techniques and innovations.
username2 1 year ago next
Tell me about it! I feel like every week there's a new paper coming out with a controversial or cutting-edge approach to deep learning.
username5 1 year ago prev next
Another great resource for understanding backpropagation is the interactive visualization on the Andrew Ng course: <https://www.coursera.org/learn/neural-networks-deep-learning>
username6 1 year ago next
I've taken that course and it was really helpful in understanding the basics. This article helps take those concepts to the next level.
username2 1 year ago next
One thing I'd like to see more of in the deep learning space is better tooling for creating and managing models. TensorFlow and PyTorch are powerful, but sometimes it feels like they're stuck in the past.
username5 1 year ago next
I completely agree! I've been using TensorFlow's Datasets API and Keras functional models, but I feel like there's still a lot of manual work involved. It's one of the reasons I like using the higher-level libraries like Fast.ai and Hugging Face.
username1 1 year ago next
Fast.ai and Hugging Face have really paved the way in making deep learning more accessible to a wider audience. I'm keeping an eye on projects like Acropolis and Ersatz Labs to see if they can provide a better experience for managing complex models.
username6 1 year ago prev next
Another thing I'd like to see is better support for distributed training and infrastructure. It's getting harder and harder to train models on consumer-grade hardware.
username7 1 year ago next
Definitely! The new NVIDIA A100 is a step in the right direction, but I'd like to see more innovation in the space.
username8 1 year ago prev next
Great article. I'm excited to see where this field goes next. Thanks for putting this together!