Day 6 – Probability Distributions for Machine Learning

Hey everyone, Dhairya here 👋

Yesterday I went through the basics of probability & statistics — mean, variance, probability rules, and distributions.
Today I went deeper into probability distributions, because these are the backbone of how ML models represent and handle uncertainty.

---

🔢 What I Learned Today

• Bernoulli Distribution – models binary outcomes (success/failure). Used in logistic regression and binary classification.

• Binomial Distribution – extends Bernoulli to multiple trials.

• Normal Distribution – the famous bell curve. Many ML algorithms assume normality in data.

• Uniform Distribution – baseline “all outcomes equally likely.”

• Why Distributions Matter in ML

  • Data preprocessing → understanding skewness, outliers

  • Model assumptions → Naive Bayes, regression errors

  • Random initialization in Neural Nets often comes from distributions (e.g., Xavier/He initialization).

---

🌱 Reflections

This was a satisfying day — distributions always felt abstract, but seeing them visualized with Python really made them click.
It’s cool to realize that “randomness” is not random at all — it follows patterns (distributions) that ML models exploit.

---

💻 Notebook

I’ve uploaded my Day 6 notebook here 👉 GitHub Link – Day 6 Notebook

---

📚 Resources

🎥 YouTube

• StatQuest – Probability Distributions

• Khan Academy – Probability Distributions

• 3Blue1Brown – Normal Distribution

🌐 Websites

• GeeksforGeeks – Probability Distributions

• Towards Data Science – Probability Distributions in ML

• Machine Learning Mastery – Probability Distributions

---

🎯 What’s Next?

For Day 7, I’ll explore Descriptive Statistics in more detail — covariance, correlation, and why they matter in ML.

See you tomorrow 👋
— Dhairya