Maybe you have a job position as a data scientist or ML researcher, but you cannot be called that way if you cannot answer the following simple questions.
- What is the minimal Artificial Neural network configuration to learn the sine function on the whole range of real values?
- How gradient descent method changes convolution kernels?
- For training purposes activation functions must be differentiable. What method we can use if the activation function is not differentiable and give an example of such function.
- How to measure amount of information stored in ANN?
- How to find synonyms with LLMs?
Answers
Might be an answer on how to train ANN on a period of the sin function the way that it can properly extrapolate outside of the training boundaries: Neural Networks Fail to Learn Periodic Functions and How to Fix It