Interactive Intro

Neural Networks

Problem

• How can we intuitively understand the neural net and apply cutting edge algorithms with intuition?

Scene 1 | Goal Supervised

Interactive Intro

Neural Networks

Definition

• You were just given a task to predict the house price given our input data- size of house in m2. Let's use industry standard Neural networks - which finds the relationship between the input and output data so that we can make predictions based on data it has never seen before.

Scene 2 | Goal Supervised

Interactive Intro

Neural Networks

Breakdown

• Show does it do that?
• Let's take a lesson from Feynman and break this complicated architecture down into simple components so that we can understand it.

Scene 3| Goal Supervised

Interactive Intro

Neural Networks

Define Neuron

Scene 2 | Goal Supervised

Our input data does not change, its the truth, we can't change the radius squared in this case. That's our x value.

Our slope which you will see as w, is what our neural network can change in order to make better predictions and learn.

Finally, the b value is our y-intercept which the neural net can change in order to make better initial starting points

Interactive Intro

Neural Networks

Illustrate equation

Scene 2 | Goal Supervised

So let's say we have a house of 5 Square meters, costing $2M and 9 Square meters costing $2.5M. What if we had 10 Square meters, what's the price? So our neural network will create a general function that will make the optimal predictions for us given input and output data.

Interactive Intro

Neural Networks

Illustrate equation

Scene 2 | Goal Supervised

So by changing our "w" value, we can see that the slope or the rise over run, changes to make better predictions.

But how good are those predictions? Our cost function will tell us the error of our predictions through the equation. 

Prediction - Actual Price. 

We then square it in order to make it absolute 

Interactive Intro

Neural Networks

Illustrate equation

Scene 2 | Goal Supervised

So let's first manually change the weight in order to make the best outcome. So we

 adjust the weights a little, adjust the bias. And, finally, we got the right outcome!

But wait, that was manual. How do you think a machine can do this automatically and learn from its mistakes using deep learning. How does your Tesla car learn, how do you automate reporting, and how do financial predictions actually work. Find out now

Chapter 1 | Goal Supervised

Interactive Intro

Neural Networks

Neuron

Chapter 1 | Goal Supervised

Interactive Intro

Neural Networks

Neuron

Chapter 1 | Goal Supervised

• But what if we had more than one input?
• Surface Area + Location?

Interactive Intro

Neural Networks

Neuron

Chapter 1 | Goal Supervised

Interactive Intro

Neural Networks

Chapter 1 | Goal Supervised

Interactive Intro

Neural Networks

Neuron and synaptic weight

Chapter 1 | Goal Supervised

• synaptic plasticity is the ability of synapses to strengthen or weaken over time, in response to increases or decreases in their activity

Interactive Intro

Neural Networks

Chapter 1 | Goal Supervised

Interactive Intro

Neural Networks

Chapter 1 | Goal Supervised

Interactive Intro

Neural Networks

Function

• A function is a block of code which only runs when it is called.

• You can pass data, known as parameters, into a function.

• A function can return data as a result.

Chapter 1 | Intro

Create a Function

To create a function, use the keyword def Name():

Example

Create a function named NeuralNet, with a property named and call it

Chapter 1 | Python, class intro

Optimization Parameters

Chapter 5 | Forward propagation Intro

Chapter | Forward propagation Intro

Chapter | Forward propagation Intro

Chapter | Sigmoid Function

OK but what is a cost function?

Chapter | Cost

Chapter | Error Calcuation

Chapter | Error Calcuation

https://www.surveymonkey.com/r/QY2W2C9

Please give us feedback :)

There are only 5 questions in the Survey