I faced the problem of choosing a neural network to work with a different number of input parameters. The input parameters are numeric values (for example, float), the number of input parameters is from 3 to 12. The goal is to create a network that will correctly process different numerical values without distorting the resul. (Preferably use Keras)
At the moment, as I understand it, there are several ways to solve this problem:
Use a fixed size of 10 and fill the empty values with zeros. However, would this be good practice, and would it distort the results too much?
Resize to a fixed size (roughly 7), but would that work for numeric values?
Using RNNs, but are RNNs mostly used for text and speech, and would they also work with numbers? Are there any other options and which ones would be the most correct? Thanks in advance for any answer or link
I need to create a neural network for my small learning project. The idea is that the user selects a certain category (e.g., a certain web vulnerability), then specifies the inherent certain characteristics of that category via input range. And then he should be shown the result of the probability of that vulnerability. Initially, there are a certain number of training examples for each individual category, but further on this will be based on the responses of the users. Perhaps there is a type of neural that is more suitable, but my teacher could not answer this question