Suppose I have a feature composed by 784 numbers, and I want to use it as input of a neural network implemented from scratch whose first layer has 64 neurons.

How can I put 784 numbers in 64 neurons?


1 Answer 1


To reduce the number of features from the raw data to 64 input neurons, you can perform dimensionality reduction. There are various methods to do this, here are some common ones that are easy to implement.

  1. Principle Component Analysis
  2. Autoencoders (basically another neural network that you can train unsupervised)
  3. Random projections (Make a 784x64 random matrix and multiply)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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