Last Updated on June 9, 2020 by Editorial Team

Author(s): Alison Sin

Neural Networks

These are the notes I’ve taken while going through this course at Brilliant.org. 🙂 since a premium membership might not be universally accessible and they did an outstanding job explaining the concept in simple terms, I decided to share some insights 😊

Note: All credits go to Brilliant.org!

I. Introduction

A. Why Artificial Neural Networks (ANNs)?

Because some problems X be solved with programming.

e.g. a vision problem: object recognition [classifying simple objects]

→ the main difference: different number of corners

B. What is an ANN?

An ANN is made up of ANs (artificial neurons).

Features of AN:

i) follow the rules mechanically from input to output

ii) learn by feedback reinforcement:

1. ANN is fed with input → make the best guess

2. if correct → nothing happens VS

If wrong → adjust the internal configuration to change computation

II. Neurons

• Types: 1. Binary Neurons, 2. Sigmoid, 3. Identity
• Act as: 1) classifiers → 2) predictors

Can adjust: 1) bias (threshold), 2) weights (influence)

1. Binary Neurons (~ Decision Box)

a) perform “AND,” “OR” by changing the bias

e.g., with two inputs (I1, I2)

if I1 + I2 ≥ bias → “ON”

b) can perform “XOR” → by introducing -ve inputs

Intro to Neural Networks (Brilliant.org) was originally published in Towards AI — Multidisciplinary Science Journal on Medium, where people are continuing the conversation by highlighting and responding to this story.

Published via Towards AI