-2

In the Deep Learning Specialization course by Dr. Andrew Ng, he mentions that stuff like ensemble models or boosting, though work fine in practice and are good for winning competitions, he recommends firmly against using them for developing real-world applications. Why is that? What's wrong with them?

Amir Valizadeh
  • 51
  • 8
  • 2
    I’m voting to close this question because it's not about programming. – bad_coder Sep 20 '21 at 00:58
  • 1
    Welcome to Stack Overflow. This question seems to be off-topic, please consider the guidelines [*"What topics can I ask about here?"*](https://stackoverflow.com/help/on-topic) – bad_coder Sep 20 '21 at 00:58
  • @AdarshWase thanks for letting me know the obvious: that I'm asking for the reason behind an opinion! – Amir Valizadeh Nov 28 '21 at 23:52

1 Answers1

1

I think you need to run multiple algos over your dataset and see which performs best, on the given data that is fed into the model. Ensemble models probably work best on most datasets, but that will certainly not be the case all the time. See the links below for some examples of how to test several models, to pick the best, based on the outcome of each, and not based on some kind of preconceived notion.

https://github.com/ASH-WICUS/Notebooks/blob/master/Accuracies%20of%20Different%20Classifiers%20-%20Wine.ipynb

https://github.com/ASH-WICUS/Notebooks/blob/master/Accuracies%20of%20Different%20Regressors%20-%20Housing%20Prices.ipynb

ASH
  • 20,759
  • 19
  • 87
  • 200