Disadvantages of Using Deep Learning

Some of the disadvantages of deep learning are:

  • They are inscrutable: you can’t understand why they have come to a particular result without having years of experience working with them and even then not always.
  • Training then and choosing the right network topology is a black art: some things work and some things don’t but it’s not clear there’s a good methodology for designing the “right” network architecture. This is especially so for some of the more exotic setups like Generative Adversarial Networks.
  • They take a lot of data (far more than other algorithms) to train.
  • They take a lot of time (far more than other algorithms) to train.
  • They take a lot of memory (far more than other algorithms) to execute. This is an issue especially on mobile.
  • You never quite know when training a neural network is “done.”
  • There’s no theoretical foundation for neural network.
  • They do not generalize well (e.g. if you train a network to drive a car on sunny days, and then test it in the rain, expect it to fail abysmally).

At DL Recruiting Partners, our experienced recruiters can help you find the temporary, contract, or direct-hire talent you need, and help you extend an offer once you have found the perfect employee. Contact us today to learn more.

Related Posts

Leave a comment