Preston Parry
Data Scientist, Machine Learning Engineer, addicted cyclist and rock climber
Conjuring Up a Neural Net
Just like picking up any other library, but sounds way cooler
Today's Tour
- What are nets good for?
- What are neurons again?
- Intro to Neural Nets
- Intro to Machine Learning in practice
- Intro to Brain.js
What Are Nets Good For
- Categorization or Continuous outputs
- Complex pattern recognition
- Example: A loan officer at a bank
- Image recognition
- Total n00bs
*Disclaimer: not actually everything, but they're a really great place to start many machine learning explorations
EVERYTHING!!*
Neurons
Neural Networks
Example Inputs and Outputs
liked
matched
age
height
religion
Example Weights
liked
matched
age
height
religion
.084
.623
.760
What does this mean?
Hidden Layers
Machine Learning in Practice
- Cleaning and Formatting Data
- Training and Testing Datasets
Training and Testing Datasets
- Make sure to test your ML algo on data it didn't already train on
-
General split
- Training: 80%
- Testing: 20%
Cleaning and Formatting Data
- Most of your time will be spent cleaning and formatting your data to get it into the format your algo expects
- Normalization
- Binarization
- Cleaning up inaccurate/incomplete data points
Brain.js
It's written in JavaScript- yum :)
Brain.js
To the Docs!!
Your turn!
Conjure up a neural net
Key Takeaways
- Machine learning does not follow the 80/20 rule
- It's even more exaggerated- 90/10
- Machine learning is just like any other type of engineering
- It works because... we ran some data through it and it works!
Conjuring Up Neural Nets in JS
By Preston Parry
Conjuring Up Neural Nets in JS
An intro to neural networks for engineers
