Neural Networks
Expectation: By the end of this lesson you should be able to answer
1.) What is a neural network?
2.) How does a neural network work?
3.) Provide two examples of a neural network working.
4.) What does a computer need in order for a neural network to work well?
5.) What can go wrong with a neural network?
Key Words to Remember: Neural Network, Input Layer, Output Layer, Hidden Layers, Training, Patterns, Prediction, Accuracy
1.) What is a neural network?
2.) How does a neural network work?
3.) Provide two examples of a neural network working.
4.) What does a computer need in order for a neural network to work well?
5.) What can go wrong with a neural network?
Key Words to Remember: Neural Network, Input Layer, Output Layer, Hidden Layers, Training, Patterns, Prediction, Accuracy
|
|
Introduction
|
|
Tesla picks up objects in the graveyard
|
Task: Answer the question:
1.) What is a neural network Consider the images up above. The AI can look at the image and identify humans, buildings and all manner of objects. Neural networks are employed to act like a brain in order to make predictions and recognize objects in mirrors. |
Neural Network Development and Image.net
AlexNetAlexNet was created in 2012 by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton at the University of Toronto. It won a big image contest called ImageNet Challenge, getting only 15.3% wrong, much better than the next best at 26.2%. AlexNet is a type of neural network with 8 layers: 5 layers to spot patterns in pictures (using special filters and tricks to speed up learning) and 3 layers to make final guesses (with a method to avoid mistakes). It used fast computer chips and extra picture tweaks to work well.
AlexNet showed everyone that deep neural networks could learn from pictures on their own, starting a boom in deep learning. It led to newer, better networks like VGG and ResNet. By June 2025, AlexNet is seen as an important starting point, and its ideas, like faster learning and avoiding errors, are still used, even though newer networks are much bigger and better. -Powered by Grok |
In 2007 a computer science professor at Stanford University started a project called "ImageNet".
The goal was to create a comprehensive dataset to advance computer vision research. At the time computer vision (the ability for a computer to recognize an object from an image) struggled as there was a very small data set. ImageNet has 14 Million images that are connected into 20,000 different categories. In order to organize these images they crowd sourced people to label the images. 
Fei-Fei Li
The team at AlexNet
|
|
|
|
Neural Networks - How they work
|
Task 1: Complete the missing words in the sentences next to each picture.
Missing Words: Input Layer Prediction Data Patterns Output layer Hidden Layers Neural Network Task 2. After completing task 1 provide an answer as to how neural networks work? |
|
1.) A _________ takes in _____ through an ________ |
|
2.)Neural network with _________ finds ________ during training |
|
|
3.) The network makes a ________ and sends it to the _________ |
UPDATE: The percentage is known as a Weighting or weight. What is the weighting that this is likely to be a dog nose in this picture.
Two Examples
For both of these tasks please answer them under: 3.) Provide two examples of an neural network working.
Give the answers in the same terms as the previous activity.
e.g. 1.)The neural network would take in the X-ray image into the input layer.
Give the answers in the same terms as the previous activity.
e.g. 1.)The neural network would take in the X-ray image into the input layer.
|
Mary has had an X-Ray scan of her chest to check for cancer.
Task: Explain how a Neural Network could use the X-Ray Image to detect cancer in the image. 
|
Steven works for the content and moderation team at a new social network called "InstaFace".
Their moderation team is inundated with blocking offensive images. Task: Explain how a neural network could help identify and block offensive images, making the website safer to browse? 
|
Problems and Solutions
Task: Read the article and answer questions 4-5 around what computers need in order for neural networks to work and what can go wrong.
https://abc7news.com/tesla-s-autopilot-self-driving-car-officials-investigating-teslas-autopilot-feature-after-fatal-crash/1410042/
In order for neural networks to be trained properly they need a large amount of data. Think about what can go wrong if a self driving car does not have a large amount of varied data.
https://abc7news.com/tesla-s-autopilot-self-driving-car-officials-investigating-teslas-autopilot-feature-after-fatal-crash/1410042/
In order for neural networks to be trained properly they need a large amount of data. Think about what can go wrong if a self driving car does not have a large amount of varied data.
Word Definitions
Now that you have completed this page, finish the definitions task at the bottom of the one note:
