I recently read a blog post by Dimitry Masin on how Monzo is using RNNs to provide improved customer support. I was wondering if the data team were strictly focusing on implementing RNNs or if other deep learning models were being looked at and for what purpose (roughly)?
If you are going to use abbreviations it would be helpfull to explain what they stand for!
RNN means a Recurrent Neural Network
How does an “RNN” work, please?
A genuine question.
What is “deep learning” please?
What is a Monzo please?
(I’m sorry. The point about abbreviations is a good one.)
At a high level, all neural networks are basically trained on input data, they learn from this and then you can use them to classify new data
I’m not current with machine learning anymore so hopefully someone else can give some more RNN specific info
So a really complex input version of predictive text?
Are you asking for the theory behind the model?
From what i can tell, let’s say you’re a customer. You perform 3 events (signup, transaction, transfer money) all these three events contain a row of information which you can transform into a vector.
You take this vector and feed it into your model which is the recurrent neural network.
What they are trying to predict is the probability of a certain issue (“transfer not showing up in mobile phone”) given your past 200 events (signup, transaction, transfer, transaction etc…).