What are the advantages and disadvantages of neural networks?

Data Science, AI and ML
,
#1

Advantages and disadvantages of neural networks.

#2

Advantages: Neural networks (specifically deep NNs) have led to performance breakthroughs for unstructured datasets such as images, audio, and video. Their incredible flexibility allows them to learn patterns that no other ML algorithm can learn.

Disadvantages: However, they require a large amount of training data to converge. It’s also difficult to pick the right architecture, and the internal “hidden” layers are incomprehensible.

#3

The advantages of deep neural networks are record-breaking accuracy on a whole range of problems including image and sound recognition, text and time series analysis, etc. There are many open source libraries that you can use to experiment with them.

The main disadvantages are:

  1. they can be hard to tune to ensure they learn well, and therefore hard to debug;
  2. they do not have explanatory power; i.e. they main extract the best signals to accurately classify and cluster data, but they will not tell you why they reached a certain conclusion;
  3. they are computationally intensive to train; i.e. you need a lot of chips and a distributed run-time to train on very large datasets.