• S1 = 60 minutes each before lunch
• S2 = 75’ before lunch
• S3, S4 = 90 minutes each after lunch

• Who we are
• What we believe
• Data Science use cases at GOJEK?
• How to spot DS opportunities?

• AI - Big Data - Machine Learning
• How all these three are related?
• Machine Learning:
  • What are different ML model classes
• Data Science Flow
• The three DS core. (Biz, Math, Coding)
• DS Problem formulation (northstar metrics to solve: BCR)
• ML 101 (Linear regression): (use Jupyter notebook for demo)
  • Parameter estimations:
    • Loss minimization (intuitively)
    • Bias-Variance trade-offs
  • Model Evaluation:
    • Briefly about model evaluation

• Intro to Case Studies 15’
• Intro to Pandas (notebook) 45’
  • Definitions
  • Basic functions
  • Merging dataframes
• Intro to Data Visualization (notebook) 45’
  • Introducing matplotlib and seaborn with some basic charts
  • A bit more not so simple charts (with interaction, sankey flow, etc)

• Pandas Exercises
• Viz Exercises

*use collabs

Keywords : Problem Definition, Correlation, IID, Practice
Gradient Descent

Keywords : Problem Definition, Information Theory, Entropy, Type of Classification, Problem, Practice (add

• 40 mins - Feature Engineering & Feature Selection

• 50 mins - Ensemble learning (bagging & boosting) - hands on

  Keywords : Averaging model, OneHot, Bootstrap, Bagging, Boosting

Keywords : Cost Function, Bias-Variance-Trade Off, Cross Validation, F1 Score, Recall, (other metrics)

• Regression: using new dataset with non-poly linear reg + plot
• Classification: using new imbalanced dataset, discuss other classification metrics, also using medium dataset (F1 score, recall, etc)
• Ensemble: boosting (adaboost) hands-on, using medium dataset(bias-variance tradeoff)

• 10 mins - Theory
• 50 mins - Guided K-Means algorithm creation (hands-on exercise)

• 60 mins - Application to K-Means on an image (hands-on exercise)
• 10 mins - Other unsupervised learning algorithms theory
• 5 mins - Homework explanation (wine data)

• 10 mins - Why an what for NN.
• 10 mins - What are computation graphs and derivations over computation graphs
• 30 mins: Hands on : Implement logistic regression in NN style.
• 10 mins: NN representation
• 5 mins: Activation functions
• 20 mins: Gradient descent and Backpropagation

• 10 mins: continue Gradient descent and Backpropagation
• 10 mins: Initialization of weights
• 30 mins: Hands on: Implement shallow NN using python and numpy
• 10 mins: Explain homework, building blocks of DNN, forward and back prop for DNN
• 5 mins: Softmax regression
• 30 mins: Hands on: Softmax regression using Keras

• S1 - Intro to the Kaggle problem, participants make group, start
• S2 - Kaggle-ing
• S4 - Presentation from the top 5 of the teams

For some module, you are required to install this environment. For the rest, we will use Google Colab

For Mac:

brew install pyenv

For Ubuntu, you need to clone the repo.

cd
git clone https://github.com/pyenv/pyenv.git ~/.pyenv

Then, modify your environment variables.

echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/<.bash_profile/.zshrc>
echo 'export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/<.bash_profile/.zshrc>
echo -e 'if command -v pyenv 1>/dev/null 2>&1; then\n  eval "$(pyenv init -)"\nfi' >> ~/<.bash_profile/.zshrc>

Note for Ubuntu, modify your ~/.bashrc instead of ~/.bash_profile.

Lastly, restart your shell so the path changes take effect.

exec "$SHELL"

• Refer to pyenv for more details

Install dependencies: (you have to be connected to Gojek Integration VPN in order to install dependencies from http://artifactory-gojek.golabs.io)

make env