Skip to content

Latest commit

 

History

History
77 lines (61 loc) · 5.34 KB

File metadata and controls

77 lines (61 loc) · 5.34 KB

Machine Learning in 6 weeks

Following is a weekly plan of learning machine learning for beginners. Resoruces and tasks have been mentioned under each week.

Week 0

For starting with python : see starting 9 lectures of https://www.youtube.com/playlist?list=PL-osiE80TeTskrapNbzXhwoFUiLCjGgY7 And start with pset 6 of https://cs50.harvard.edu/x/2020/psets/6/

Week 1

Numpy task and Resources: task 01 available at https://github.com/MananSoni42/ML-SIG-2019

Week 2

Linear Regression for single variable from scratch using numpy Resources : The tutorials at : https://pythonprogramming.net/machine-learning-tutorial-python-introduction/ All parts From “Practical Machine Learning tutorial with python - Introduction” to “ Regression - How to program R squared” Dataset : https://github.com/ashryaagr/acm_task_dataset . Download this and start working. Use regression and train using the data in train.csv and then compute accuracy using data in test.csv

Week 3

Use sklearn SVM on NASA asteroid dataset and train it to predict the property “hazardous” (last column of data ) of asteroids. You have to only use the data available in nasa.csv file. Resources : The tutorials at : https://pythonprogramming.net/machine-learning-python-sklearn-intro/ Some knowledge of Pandas will be required ..so go through these tutorials : https://pythonprogramming.net/introduction-python3-pandas-data-analysis/

Challenges that you will face : Some columns are strings. You need to convert them to some integer. ( Try to think of some way to do that ) Dates are given in various formats. Find some way to convert them to integers to feed them to model. Some fields ( columns ) are useless ( standard term for this is feature selection ). It might be possible that a field (column ) which contains strings is also useless. Dataset : https://www.kaggle.com/shrutimehta/nasa-asteroids-classification

Week 4

https://github.com/ashryaagr/ML_Recruitment You have to fill the Todos of neural_net.py. Resources are provided in Readme. You can skip the task 1. After completing the Todos, use this Neural Network on MNIST dataset to predict the digit in images

Week 5

Now You have to make a Generative Adverserial Network using the Neural network you just made

This time you yourself will have to search for resources... But some suggestions are Research papers, lecture on GAN in the series CS231N, tutorial by lan goodfellow ( who is the creator/father of Generative adverserial networks)..you guys can also use the tutoial he presented in NIPS.. Or there are a number of videos on youtube.. Or on tensorflow website in tutorials, they make GAN using tensorflow (but you guys have to make it using the Neural network you wrote)

Btw you might have to rewrite cost function and prepare 2 separate cost functions.. One for generator and one for discriminator

Deep Learning and cool stuff

Machine Learning Resource List

ML Winter SIGs