REVIEW 5

The questions that follow are based on

1. Tools: Scikit-Learn
2. Tools: NLTK

Scikit Learn

• Why is it a good idea to shuffle your training data?
• When could shuffling be a bad idea? Give 2 examples.
• What is a Binary classifier? How is it different from an n-ary classifier?
• What is a confusion matrix
• Write a snippet to obtain the confusion matrix from Scikit learn
• In the above confusion matrix that you obtained from scikit-learn, what do each of the cells represent. In other words, fill the following matrix:

--|-------------|-------------------|------------------|--
  |             |        ?=?        |        ?=?       |
--|-------------|-------------------|------------------|--
  |     ?=?     |         ?         |         ?        |
--|-------------|-------------------|------------------|--
  |     ?=?     |         ?         |         ?        |
--|-------------|-------------------|------------------|--

• With the above confusion matrix, write a snippet to compute:
  • Precision
  • Recall
  • False Alarm
  • F1 score
• What is a k-fold cross validation?
• What is stratification? Why is used?
• When would k-fold cross-validation be misleading?
• How would one convert a linear regression into a binary classifier?
• What is a decision hyperplane?
• Say you have an n-dimensional data, how many dimensions would the decision hyperplane comprise of?
• What is the key intuition in SVM?
• What does an SVM aim to achieve when training?
• SVM seem to work very well for which domain? (Ans: Text Classification)
• What are hyper-parameters?
• What is the difference between CART and Random Forest?
• How does Random Forests work?
• What is a mode of an array?

Text Classification

• Draw a general pipeline for a simple text classification task.
• Define in a sentence or two, what each of the blocks above aim to achieve.
• Define the following terms:
  • Tokenization
  • Lemmatization
  • Stopword removal
  • Part-of-speech tagging.
• Given a sentence below, convert it into tokens using: Tokenization, Lemmatization, and Stopword removal.
  • "This is a sentence about cats"
• What is vectorization?
• Give the mathematical formulation of the following:
  • Term Frequency
  • Document Frequency
• What is intuition underlying TFIDF vectorization?
• Generate a TFIDF vectorization of the following two sentences:
  1. "Something about a cat."
  2. "Something about a dog."
• How does sk-learn store sparse matrices? (Ans: CSR Format)
• What is the CSR (compressed sparse row) format?

Multiprocessing

• What is Global Interpreter Lock (GIL)?
• What is chunking?
• Distinguish between synchronous and asynchronous processes.
• How would you count the number of processors in your system with python's Multiprocessing?
• Write a snippet to print "Hello, World! {processor id}" on each of the systems processors.