I am building a feed-forward neural network with two hidden layers, which I will train with a medical dataset, which consists of both data, such as age and sex, and images of x-ray scans ($1024 \times 1024$). The labels are types of cancer.

I believe that ages and sex will affect the output of the network. But including the images will make the network biased towards the images, because images will occupy most of the input layer.

How can I design the input layer of the network?

  • $\begingroup$ @BrianO'Donnell A neural network with two hidden layer is a deep neural network $\endgroup$ – kkr4k Nov 29 at 7:05
  • $\begingroup$ @kkr4k No expert in neural networks considers two hidden layers as a 'deep neural network. nbro corrected the original erroneous reference to 'deep neural network'. $\endgroup$ – Brian O'Donnell Nov 29 at 19:45

I would give some advice here:

You should probably use Convolutional Neural Network for image data or you'll end up with too many input features, and too much computation. What are your constraints on this point?

I would not mix image data with numerical data. Maybe you can implement two neural networks, one handling image data and another managing the numerical data, each one outputting their results. A second step would be to use their results to handcraft a global result or let another NN manage this.


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.