7 PM – AI/ML – Darin

Today’s Activities:

1. Continued multi-class classification on the MNIST dataset.
2. Learned about data augmentation through the batch_tfms parameter for the dataloader.

Homework:

For Reine:

1. Create a google doc under your homework google drive 
2. Write up your project proposal for training your own AI model detailing:
   – What you will be classifying (has to be multi-class classification!)
   – Where your data comes from (provide a link or methodology of getting the dataset). If you are using an existing dataset, you have to add more images from online means (gather data from outside of kaggle, whether it’s online images, etc)
   – How you will train your AI model (what transformations on the data you will use, and the flow of training)
   You can view the list of transformations here: https://docs.fast.ai/vision.augment.html#aug_transforms

Ensure the google doc is public access and that it is in your homework folder.

For everyone:

1. Continue working on your final project. You want to at least ensure you have the data loader setup with the proper transformations (also at this point it should be easy to run training once you have the dataloader). Upload your latest progress as an ipynb to the google drive.
2. Create a new kaggle notebook Mar9_PandasHW.ipynb and do the following:
   Youtube.csv 
   1. Use this csv file – https://drive.google.com/file/d/1kP6A9y0UBssOg3Exunv9Mnmilb0657Sh/view?usp=drive_link 
   2. Load the data
      
   3. Show only the Channel and Subscribers columns
      
   4. Find channels with more than 2000 subscribers
      
   5. Add a column Subs_per_Video
      
   6. Which channel is the most efficient?

HW Note: you can display rows with conditions like this:

df_math_greater_than_80 = df[df["Math"] > 80]
print("\nStudents with Math Score Greater Than 80:")
print(df_math_greater_than_80)
print(df.head())

Notes:

You can reach me at ddjapri@ayclogic.com.

All class notes can be found here.