When are Neural Networks applicable in Decision Intelligence?

  • 13 April 2022
  • 1 reply
  • 66 views

Userlevel 2
Badge

Are neural network or deep networks inherently better than classical machine learning algorithms?

 

Before joining Peak as a Data Scientist, my answer to that would be definitely yes, 100%. That opinion was shaped by the fact that my previous experience was only in deep neural networks and only learnt about classical machine learning algorithm in my Masters in Machine Learning. This was also heavily influenced by the fact that the news I read on machine learning advancements were about AlphaGo, DeepFakes, and Natural Language Processing. I thought neural network algorithm was king, even used it in my data challenge for the Graduate Scheme Assessment day. 

 

Since starting at Peak, I have learnt that neural network are not all that in the current landscape of Decision Intelligence. Many of my colleagues have built incredible solutions for customers from Inventory Optimisation, Price Optimisation, Forecasting Demand, and much more. Some have even tried to implement deep neural network algorithm which do not always outperform the classical algorithms. Even when it does outperform classical algorithms, there are some questions we need to ask ourselves :

  • How much time was spent to get that 1-2% increase in accuracy?
  • Is the neural network algorithm you built explainable to a non-technical person?

“Baseline can be hard to beat”

 Mark Douthwaite (Data Science Team Lead, Peak Labs)

 

With all being said, I do not think the future is bleak for deep neural networks in Decision Intelligence. There are certain specific use cases of deep neural network algorithms that are known to be better than classical algorithms, most notably Image Processing/Classification/Segmentation and Natural Language Processing. If we find a way to leverage those algorithms to further enhance what we can do with Decision Intelligence, there will be a need for neural networks.

 

To end my post, I have a question of discussion, when do you think neural networks is applicable in Decision Intelligence and do you agree/disagree with my opinion on neural networks?


1 reply

Userlevel 3
Badge +1

Important disclaimer: I’ve never worked with a neural network on a real problem!

For me, I would probably consider trying out a NN / some kind of deep learning if:

  • I had a lot of data and
  • I had a lot of development time and
  • The relationship I was trying to model was highly nonlinear and complicated and
  • It’s not immediately clear what the features should be and
  • Small percentage points of accuracy would deliver an appreciable increase in value (whatever that might be).

I am aware they can be developed relatively quickly if you know what you’re doing, but I don’t.

Most of the benefit seems (to me, a NN non-expert) to come from the “automatic” feature extraction that the network does.

Reply