Project2 of #100MLProjects — Classifier: MNIST Digit Recognition
How to eat an Elephant? One bite at a time.
About #100MLProjects:
100MLProjects is a Challenge where I try to attain proficiency in Machine Learning and Deep Learning concepts by doing 100 Projects. The complexity of the projects keeps increasing as I progress through the challenge, so other Machine Learning/Deep Learning aspirants can also follow this path.
Recap:
I chose ‘Graduate Admissions Prediction’ as the first of my 100 project challenge. This project uses different basic Machine Learning Algorithms and tries to predict the chance of getting an admit in Universities for the given student profile. I have also compared the performances of different algorithms, and visualized which features are playing a crucial role in the admissions based on the dataset(Read more here: #100MLChallenge Graduate Admissions Prediction).
Note: Thank you, Thameem Abbas, Aiswarya, KeerthiVasini, ArvinthRaaj, Yatheen, Gokul Diwakar, Venkat, and others for your words of appreciation, suggestions, and feedback. I highly value your inputs and will incorporate your feedback into all my upcoming projects.
Previous Feedback: To Share the project idea before starting the projects, so other interested people could also join me. (So here I’m! Posting Pre-Project Update)
Pre-Project Update: Project #2 MNIST Digit Classifier
For the second project, I decided to do go with the classic Machine Learning Hello world problem — Handwritten digit classification using the MNIST Dataset(Project Suggestion: Thameem Abbas). The first project was a regression problem, and this is a classification problem.
MNIST Handwritten Digit Classification Dataset comprises of 60K grayscale images of handwritten numbers. (Dataset available here).
The Goal of this project is to “Build a Classifier model that predicts a handwritten number”.
Since we are starting our Machine Learning journey, I’m refraining myself from using advanced complicated algorithms like CNN’s. I’m planning to use other Machine Learning algorithms like K-Nearest Neighbors, Support Vector Machines(SVM), and Neural Networks(NN) to build Classifiers.
Also, This project will include a GUI. The canvas in GUI allows the user to draw a digit, and the classifier will try and predict the number drawn by the user.
If you are interested, comment below or ping me on LinkedIn(Link below). Let’s make our own versions of this project, and let's share our lessons, insights, tips, and tricks that we learned through this course of this project.
If you are also willing to take up this project, here are some useful resources that can help you get started. (Note: I’m going to look-up at very few resources, and build most of the project on my own, so I could get a better understanding)!
- Dataset: http://yann.lecun.com/exdb/mnist/ (or) Kaggle: https://www.kaggle.com/c/digit-recognizer/data
- Himanshu Beniwal’s Blog Article: Handwritten Digit Recognition
I will share the Post-Project update soon!
Until then! Happy Learning! Happy Machine Learning!
-#Laxmena
#StaySafe
