What are the three core types of deep learning?
I am currently taking an intro course and the instructor mentioned three main types of deep learning architectures, but I think I missed the distinction in my notes. Are we talking about supervised, u...
What makes transformers better than RNNs?
I have been studying sequential models and I am trying to figure out why everyone moved to Transformers. If RNNs are designed for sequences, why are they being replaced? I am particularly interested i...
What are some best practices for model serialization?
I have trained a successful model, but I keep running into issues when I try to save it and load it back up in a different environment. I am getting mismatched shape errors. What is the standard way t...
What is the role of the bias term in a neuron?
I was reviewing the basic equation for a neuron, and I see the weights and the inputs, but the bias term is still a bit abstract to me. Does it just shift the activation function? If I remove the bias...
How do I interpret my training logs?
I have a bunch of charts in TensorBoard showing training loss, validation loss, and accuracy, but I am not always sure what they are telling me. What should I be looking for to detect early signs of t...
How much math do I really need for deep learning?
I am a software engineer trying to pivot into machine learning, but the math requirements are intimidating. Do I need to be a calculus expert to be effective in this field? Specifically, I am wonderin...
What is the best way to handle 'noisy' labels?
My dataset was scraped from the web, and there is definitely some human error in the labeling. I am worried that my deep learning model will memorize these errors. Is there a way to make the model mor...
What is the best way to handle sequence lengths in RNNs?
I am dealing with variable length sequences for an NLP task. Do I just pad everything to the maximum length, or is there a more efficient way? I am worried about wasting computation on padding tokens....
How do I explain deep learning to non-technical stakeholders?
I have to present a project proposal to my manager, but they are not a technical person. They keep asking how the 'brain' of the model works. What are some good analogies for explaining backpropagatio...
Does normalization really speed up training?
I have read that normalizing inputs is standard, but does it make that big of a difference? I am curious if batch normalization serves the same purpose as input normalization. I have been skipping it ...
