0
$\begingroup$

Neural networks can have multiple inputs. But some times two or more of these inputs can often be related to a single entity.

E.g : Height and weight of a person to predict the probability of disease or price and Volume of a stock to predict it's value.

How can we make a neural network understand that two inputs are related to the same entity? Is there any efficient (or inefficient) way of 'tagging' them together?

$\endgroup$
  • $\begingroup$ What problem are you trying to solve? Why do you want to do this? $\endgroup$ – D.W. Dec 18 '13 at 8:55
  • $\begingroup$ I am trying to make a neural network to predict stock market crashes, my particular problem is related to the price and volume of a stock. Feeding both as different unrelated inputs seems to be counterproductive. But I was hoping for a generic answer... as the same problem might appear in other scenarios $\endgroup$ – Shayan RC Dec 18 '13 at 9:48
  • 1
    $\begingroup$ I'm not sure why you think that feeding them as inputs will be counterproductive. Normally you take the independent variables and feed them as inputs, and train the model to try to predict the dependent variables. Sometimes (especially with neural networks) it can be helpful to preprocess the inputs, e.g., to normalize them (subtract the mean and divide by the standard deviation) and to reduce correlation between them (e.g., using PCA or some other form of decorrelation). I recommend you read some textbooks and tutorials on machine learning and neural networks. $\endgroup$ – D.W. Dec 18 '13 at 22:36
  • $\begingroup$ Please forget that I said that it might be counterproductive, all I mean is that it might be more intuitive to make the neural network understand that the numbers I'm feeding it as input are different aspects/properties of the same entity. Especially when the inputs are important to the output. P.S. I am already normalizing the inputs and thanks for mentioning PCA, I'm already looking into how I can apply it here... $\endgroup$ – Shayan RC Dec 19 '13 at 3:35
1
$\begingroup$

It is not necessary to indicate this to the neural network (and not possible). This is not part of how neural networks are typically used -- or part of how other machine learning algorithms are used, either.

I confess I'm not clear on why you think it would make a difference, but it doesn't matter: there's no way (that I know of) to do that, and it's not something people normally do.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.