Writing Our First Classifier - Machine Learning Recipes #5

Uncover the intricacies of creating our first classifier from the ground up, making this tutorial an essential milestone for newcomers to the field. In 8 comprehensive steps, we'll guide you through the process of writing the classifier, starting with the code introduced in episode #4 and comment out the classifier we imported. Then, we'll code up a simple replacement - using a scrappy version of k-Nearest Neighbors. Explore the pros and cons of this approach as you dive deeper into understanding the mechanics of writing a classifier. Don't miss this opportunity to expand your knowledge and master the art of crafting classifiers from scratch!

Chapters:
0:00 - Intro
0:00 - Outline
0:56 - Step 1: Comment out imports
1:22 - Step 2: Implement a class
1:38 - Step 3: Understand the interface
2:08 - Step 4: Get pipeline working
3:14 - Step 5: Intro to k-NN
4:38 - Step 6: Measure distance
6:13 - Step 7: Implement nearest neighbor algorithm
7:22 - Step 8: Run pipeline
7:47 - Pros and Cons for this algorithm
8:24 - Wrap up

#ML #machinelearning

[Abdullah-mgzl]

*quick summary:*

- recall, a classifier is first trained (where it receives a bunch of training inputs along with the corresponding outputs) and then it is asked to predict the output for new inputs.
- one basic way of implementing a classifier is to save all of your training input/output pairs, and then when you are asked to predict the output for a new input, return the output of the closest training input to your new input

- so for example, let's say I give you the following training inputs (1, 12, 130) and their respective outputs (2, 14, 140).
- and then I ask you to predict the output for a new input, 4
- well, 4 is closest to 1, and the output for 1 was 2, so we will predict the output for 4 to be 2 as well

- the inputs in the above example are 1 dimensional, but you can use the distance formula for 2, 3 or any number of dimensions

Thanks again for the videos, very clear and helpful!

[EmettSpeer]

I subscribed to Google Developers just to be notified about your videos.

Please release your videos more often. I personally have found your explanations and examples the best way to get started in this topic.

[shutupntakemymoni]

Just when I thought this world is filled with people who loves to make things look complex to scare beginners away, here is a savior who can make things easy to understand and appreciate for beginners. Thank you so much!

[xavdest]

Is this guy a leader in the field? He explains it so well, and if you can explain it so everyone understands it, that means you really know it.

[yashpundir]

Oh Boy, the happiness I found after writing my first classifier and perfectly understanding it, was ENORMOUS. Thanks for these great videos.

[armitra]

Oh God! Why did I not see your videos before? You make these concepts so much more accessible, and here I was struggling so much before.

[ygpsk]

this is the most useful set of ML tutorials i've seen online, and i've seen a lot! Thank you so much for making these!

[Radiokot]

This man is so happy, I cant stop smiling c:

[ltluvideo]

This is the best ML tutorial I have seen so far. Most of them have too much math, not enough code.
Please keep it up! We want more.

[jpggvilaca]

best tutorial out there about machine learning. Simple and concise, just perfect.

[arvindroshaan]

Hi Josh and Team, This sequence makes ML fun and interesting than other online lectures. Keep it up! Keep moving forward! Your work inspires a lot of people to work in ML.

[FacadeMan]

By far, the simplest explanation for a lot of otherwise scary concepts.

[SamBeera]

Thank you for the crisp and concise video Josh, the material and your presentation is awesome !

[WilloftheWinds]

It really feels like an accomplishment after understanding this, I am so happy. Thanks for the great video Josh, keep up the good work!

[yosaikan]

Great video, very clear and very instructive.
Thanks Josh.

[TheSot]

This is great and helps to understand how simple ML algoritms work. Thanks!

[pavelorlov]

Subscribed almost immediately. Thanks a lot for your brilliant ELI5 style. Just one question? Why is X always a capital and y is always lowercase in your examples?

[lucasklaassen]

Josh, thank you for continuing to make these videos!

[homer]

I've been following along using Python 3.5 While I'm no expert in Python I did take a mild course from codecademy so this was easier to follow for me. However I still run into the occasional syntax error because of the changes between versions. Such as print('string') instead of print 'string' . Also make sure you import random or random.choice will throw an error "name 'random' not defined". Personally I enjoy translating as much as practicing these examples. If you want a quick primer on python 3(.42) look up Python Programming by Derek Banas. He has a great video that covers most of python in around 40 minutes.

[harolddavies]

Hi Josh, I love your series, it is very clear. Would you mind posting the code on Github or a link so that people can walk through the code themselves? Typing it out by hand would really help some of us. Thanks