Probability

probabilty formula

Probability is one of the 3 core pillars of Machine Learning (Linear Algebra + Calculus + Probability).

Probability is the measure of how likely an event is to happen.

It always ranges from 0 to 1.

• 0 → Impossible
• 1 → Certain
• 0.5 → 50% chance

Intuition: A bag of balls

Imagine a bag with 5 balls.

Suppose:

• Red = 1
• Blue = 1
• Green = 1
• Yellow = 1
• Purple = 1

Question:

What is the probability of picking the red ball?

There is 1 favorable outcome out of 5 total outcomes.

Answer = 20%

Probability Formula

The fundamental formula is:

Where:

Symbol	Meaning
P(E)	Probability of event E
Favorable	Outcomes you want
Total	All possible outcomes

This is the most important probability formula.

Example 1: Dice

A standard dice has 6 faces.

Question:

Probability of getting 4?

≈ 0.167 or 16.7%

Example 2: Coin Toss

Possible outcomes:

• Heads
• Tails

Probability of Heads:

50%

Probability Scale

Probability	Meaning
0	Impossible
0.2	Unlikely
0.5	50–50 chance
0.8	Very likely
1	Guaranteed

ML Use Case 1: Spam Detection

An email arrives.

AI predicts:

Class	Probability
Spam	0.92
Not spam	0.08

Interpretation:

• 92% chance it's spam
• 8% chance it's genuine

The model outputs probabilities, not certainty.

ML Use Case 2: House Price Classification

Your real estate AI predicts:

Class	Probability
Budget	0.10
Premium	0.75
Luxury	0.15

The highest probability is Premium (75%).

The prediction is based on probability.

ML Use Case 3: Face Recognition

A camera detects a face.

Person	Probability
Alice	0.95
Bob	0.03
Charlie	0.02

AI chooses Alice because she has the highest probability.

ML Use Case 4: ChatGPT

When ChatGPT predicts the next word, it assigns probabilities.

Example prompt:

"Machine learning is..."

Word	Probability
powerful	0.41
useful	0.30
amazing	0.18
difficult	0.11

ChatGPT samples from this probability distribution to generate text.

Python Example

# Probability of drawing a red ball

favorable = 1
total = 5

probability = favorable / total

print(probability)

Output:

0.2

Key Terms

Term	Meaning
Experiment	Tossing a coin
Outcome	Heads
Event	Getting Heads
Sample Space	All possible outcomes
Probability	Likelihood of the event

Sample Space Example

Dice:

Coin:

Summary

Concept	Formula
Probability	Favorable / Total
Coin	1/2
Dice	1/6
Range	0 ≤ P ≤ 1

Memory Trick

Probability = Number you want ÷ Total possible outcomes

Interview Question

Q: Why is probability important in Machine Learning?

Answer: ML models rarely output absolute certainty. Instead, they estimate the likelihood of different outcomes—such as spam vs. not spam, disease vs. healthy, or the next word in a sentence—using probabilities between 0 and 1.

conditional probabaility

Conditional probability is one of the most important probability concepts in AI and Machine Learning. It is used in spam filters, medical diagnosis, recommendation systems, and Bayesian models.

Conditional Probability = The probability of Event A happening, given that Event B has already happened.

The keyword is "given".

Real-life intuition: Property buyers

Suppose your real estate platform has 100 users.

• 40 are Premium Members
• 25 are Premium Members who booked a site visit

The question is not:

What's the probability a random user booked a visit?

Instead, it's:

Given the user is Premium, what's the probability they booked a visit?

That is conditional probability.

The formula

Example with sample values:

Where:

Symbol	Meaning
P(A	B)
P(A ∩ B)	Both A and B happen
P(B)	Event B has happened

A ∩ B is the part of B where A also happens.

Read it as:

Probability of A, given B

Example 1: Premium users

Given:

Group	Users
Total users	100
Premium users	40
Premium + Site Visit	25

We want:

Step 1:

Favorable = 25

Step 2:

Given Premium = 40

Answer = 62.5%

Meaning:

If a user is Premium, there is a 62.5% chance they book a site visit.

Why do we divide by Premium only?

This is the biggest confusion.

Imagine 100 students.

Only 20 wear glasses.

Question:

Given the student wears glasses, what's the chance they are left-handed?

You ignore the other 80 students.

Your world becomes only the people with glasses.

Conditional probability changes the sample space.

Original sample space: 100 students

New sample space: 20 students

That's exactly what "given" means.

Example 2: Dice

Roll one dice.

Question:

Given the number is even, what's the probability it's 4?

Sample space becomes:

Only 3 outcomes remain.

Favorable:

Answer = 33.3%

Notice we did not divide by 6.

We divided by 3.

Example 3: Rain

Suppose:

Event	Probability
Rain	0.30
Carry Umbrella	0.20
Rain & Umbrella	0.18

Question:

Probability of carrying an umbrella given it rains?

Answer = 60%

ML Use Case 1: Spam Detection

AI predicts spam using conditional probability.

Email contains the word:

"FREE"

Historical data:

Condition	Count
Emails with FREE	500
Spam + FREE	450

Question:

What's the probability an email is spam given it contains FREE?

90%

This is the foundation of the Naive Bayes classifier.

ML Use Case 2: Medical Diagnosis

A patient tested positive.

Data:

Condition	Count
Positive tests	100
Actually sick + Positive	80

Question:

Probability patient is sick given positive test?

80%

Notice this is different from overall disease probability.

ML Use Case 3: Recommendation System

Your property app.

A user searched:

• 3 BHK
• Whitefield
• Under ₹1 Cr

Historical data:

Event	Users
Same search	200
Same search + Contacted Builder	120

Conditional probability:

60%

The platform can predict purchase intent.

ML Use Case 4: ChatGPT

When generating text, GPT doesn't ask:

What's the probability of the word "learning"?

It asks:

Given the previous words, what's the probability of the next word?

Example:

Prompt:

Machine ______

Next Word	Probability
Learning	0.70
Vision	0.12
Translation	0.08
Robot	0.10

This is conditional probability:

The previous words become the condition.

Python Example

# Premium users example

premium = 40
visit_and_premium = 25

p = visit_and_premium / premium

print(p)

Output:

0.625

Normal vs Conditional Probability

Normal Probability	Conditional Probability
All outcomes	Given a condition
Divide by total	Divide by the given event
P(A)	P(A
Random user	Premium user only

Summary

Concept	Formula
Probability	Favorable / Total
Conditional Probability	P(A ∩ B) / P(B)
Symbol	P(A
Meaning	Probability of A given B

Memory Trick

Probability asks: "How likely is A?"

Conditional Probability asks: "How likely is A, given B has already happened?"

Interview Question

Q: Why is conditional probability important in Machine Learning?

Answer: Most ML predictions depend on existing information. Spam filters predict spam given the email content, recommendation systems predict user actions given browsing history, and language models predict the next word given previous words. Conditional probability models these "given" relationships.

Bayes Theorom

Bayes' Theorem is one of the most important mathematical foundations of AI. It is used in spam detection, medical diagnosis, search ranking, recommendation systems, and Bayesian Machine Learning.

Bayes' Theorem updates an old probability using new evidence.

Think of it as:

Old belief + New evidence = Updated belief

Real-life intuition: Medical test

Imagine a hospital.

A patient gets a positive test result.

The question is not:

Is the test accurate?

The real question is:

Given the test is positive, what is the probability the patient actually has the disease?

That is exactly what Bayes' Theorem answers.

Bayes' Theorem formula

Read it as:

Probability of A given B = (Probability of B given A × Probability of A) ÷ Probability of B

Symbol	Meaning
P(A	B)
P(B	A)
P(A)	Prior probability
P(B)	Evidence

Understanding each term (spoon-fed)

Imagine 10,000 people.

Suppose:

• 200 people have the disease
• Test detects 170 correctly
• Total positive tests = 500

Prior Probability

Before seeing the test:

2%

This is your initial belief.

Likelihood

Among sick people:

170 test positive.

85%

This measures how good the test is.

Evidence

Overall positive tests:

500

5%

Solve with Bayes' Theorem

We want:

Apply Bayes:

Step-by-step:

Answer = 34%

Important insight

Even with an 85% accurate test, a positive result does not mean an 85% chance of disease.

It is only 34% because the disease is rare.

This is why Bayes' Theorem is so powerful.

Visual intuition

Think of probabilities flowing through evidence.

Bayes updates the belief after seeing evidence.

ML Use Case 1: Spam Detection

An email contains:

WIN FREE MONEY

Historical data:

Event	Value
Spam emails	20%
FREE appears in spam	90%
FREE appears overall	30%

Bayes computes:

Meaning:

Given the word FREE appears, what's the probability the email is spam?

This is the core idea behind the Naive Bayes classifier.

ML Use Case 2: Medical AI

An AI predicts cancer.

Input:

• MRI
• Blood report
• Age

Output:

The AI continuously updates probability as new evidence arrives.

ML Use Case 3: Recommendation System

Your property app.

A user:

• Searches Whitefield
• Budget ₹80L
• 3 BHK

AI asks:

Given these behaviors, what's the probability they'll book a visit?

This helps prioritize leads.

ML Use Case 4: ChatGPT

Suppose the prompt is:

Machine

Possible next words:

Word	Prior
Learning	0.30
Vision	0.20
Robot	0.10

Now the next context is:

Machine Deep

Bayes-style updating increases the probability of Learning while decreasing others.

Modern transformers don't explicitly implement Bayes' formula, but they perform the same high-level task: updating probabilities using context.

Bayes vs Conditional Probability

Conditional Probability	Bayes' Theorem
Finds P(A	B)
Direct relationship	Updates belief with evidence
Uses existing data	Combines prior + likelihood
Example: Visit given Premium	Spam given FREE

Think of it like this:

• Conditional Probability asks a question.
• Bayes' Theorem gives a systematic way to answer it.

Python Example

# Bayes Theorem

prior = 0.02
likelihood = 0.85
evidence = 0.05

posterior = (likelihood * prior) / evidence

print(posterior)

Output:

0.34

The updated probability is 34%.

Summary Table

Term	Meaning
Prior	Initial belief
Likelihood	Probability of evidence if A is true
Evidence	Overall probability of evidence
Posterior	Updated belief after evidence

Memory Trick

Prior → What I believed before.

Evidence → New information arrives.

Posterior → What I believe now.

Interview Question

Q: Why is Bayes' Theorem important in Machine Learning?

Answer: Bayes' Theorem allows ML systems to update predictions as new evidence becomes available. It is fundamental to Naive Bayes classifiers, probabilistic reasoning, medical diagnosis, spam filtering, recommendation systems, and uncertainty-aware AI models.