• December 01, 2023, 01:11:07 AM
  • Welcome, Guest
Please login or register.

Login with username, password and session length
Advanced search  


This Forum Beta is ONLY for registered owners of D-Link products in the USA for which we have created boards at this time.

Author Topic: Yet another neural network question  (Read 2807 times)


  • Level 3 Member
  • ***
  • Posts: 103
Yet another neural network question
« on: January 06, 2023, 03:51:36 AM »

I had two questions over neural networks....

How do neurons form paths? I understand that all a neuron does is multiply every weight with their own output, find the summation of all those values and subject it to a threshold, but how do paths form? If every neuron got the same inputs from the previous layer, how does it "do better" or pass the threshold better than other neurons if they all get the same inputs deep fake app? For example, there's an input layer of x1, x2 and x3, all passing the threshold. Next layer down the line has four neurons, all of them getting the summation. How do certain neurons in that second layer form better paths when every neuron is connected to every single one in the next layer?

How do neural networks "remember" pathways and store values? Thanks for any help.