Machine learning models can lie if you're not careful.

Here's how🧵 👇🏻
Not all machine learning models are the same but typically we look at 2
metrics for determining how good a model is.

They are (generally):
A. Accuracy : How accurate our model is (more the better)
B. Loss : How 'wrong' our model is (less the bettter)
In order to avoid overfitting, we split our data into 2 parts, one for training the model and the other for 'Validation'.

After every round of training the model,it is evaluated on the validation set.

This is done to ensure that the model does not memorize but learns instead.
Overfitting is a topic for another thread but accuracy and loss are common metrics for determining how good are model is.

There are several cases where accuracy and loss can paint a false picture for us.

Here's an example 👇🏻
(Taken from this thread)
Let's say you want to train a model to predict the occurance of a car crash given certain parameters.

This silly model below is 99% accurate because 99% of the time car crashes don't happen!

This obviously isn't what want, our model 'lies' to us.
In this situation we need to optimise our model for a different evaluation metric because using accuracy or loss will not help us.

In the scikit learn documentation, there are a bunch of these metrics which are used in different situations.

scikit-learn.org/stable/modules…
As a part of my research project with Aicrowd in the next couple of months, I will be taking a look at these evaluation metrics in detail.

My goal is to make a wiki of some sort which anyone can refer for learning about these metrics in simple way as per their use case.
tl;dr

Accuracy and Loss aren't the only metrics that are out there, and if you're not careful your model can give you results that you weren't looking for.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Pratham

Pratham Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @PrasoonPratham

3 May
There are a million machine learning tutorials and courses out there which makes it really hard to choose the right one for you.

Here are some that I really like that you might find useful.
🧵 👇🏻
Before we begin, here's some unsolicited advice.

It is not about how many courses you take that matters but the problems that you solve with your knowledge.

You can take any combination of courses and tutorials from this list that you think will be useful.
Python crash course by Traversy Media (YouTube)

• Introduces you to a lot of basic concepts in Python
• Only 1.5 hours long
Read 14 tweets
19 Apr
This is a step-by-step guide on how you can solve the Titanic disaster challenge on Kaggle.

🧵 👇🏻 Image
Kaggle challenges are a great way to practice your machine learning skills.

In this thread, we'll go through each step for solving the beginner friendly titanic disaster challenge.
These are the key steps that we will go over:

- Cleaning the data
- Training a machine learning model using decision trees in Sklearn
- Making a submission to Kaggle using the predictions from our training model
Read 28 tweets
17 Apr
This is a complete roadmap for mastering Python.

🧵 👇🏻
Basic Topics

- Variables
- Conditions
- Chained Conditionals
- Operators
- Control Flow (If/Else)
- Loops and Iterables
- Basic Data Structures
- Functions
- Mutable vs Immutable
- Common Methods
- File IO
Intermediate Topics

- Object Oriented Programming
- Data Structures
- Comprehensions
- Lambda Functions
- Map, Filter
- Collections
- *args & **kwargs
- Inheritance
- Dunder Methods
- PIP
- Environments
- Modules
- Async IO
Read 6 tweets
16 Apr
I've spent countless hours trying out different themes, fonts and what not to make VS code look beautiful.

Here's how you can do same in 5 minutes.
🧵 👇🏻
The Theme: One Dark Pro

This theme is very similar to what is used in the Atom text editor and is my theme of choice.

Looks great on Python and JavaScript.

marketplace.visualstudio.com/items?itemName…
Icons: Material Icon Theme

This extention adds icons to the files, a nice small detail.

marketplace.visualstudio.com/items?itemName…
Read 6 tweets
15 Apr
A list of my favourite tutorials for learning Python as a beginner.

🧵 👇🏻
All the tutorials below include the basics like installation, variables etc.

I've also listed out the key highlights of each tutorial so that it is easy for you decide which one to pick.
Before going through these tutorials I would highly suggest you to go through this thread if you are a complete beginner.

Read 10 tweets
28 Mar
I am on Bitclout, buy $prasoonpratham before we go to the moon 😉

bitclout.com/u/prasoonprath…
Wow 2k in 8 minutes! 🚀
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!