Probability and Statistics for AI

Probability and Statistics for AI

Life is uncertain. Will it rain tomorrow? Will a customer buy our product? Will a disease test be positive? AI models quantify uncertainty using probability and statistics. This chapter teaches the math of uncertainty.

Part 1: Probability Basics

Definition: Probability of event A = (favorable outcomes) / (total possible outcomes).

# Example: Probability of rolling a 6 on a die
P(6) = 1 / 6 ≈ 0.167

# Probability of rolling even (2, 4, 6)
P(even) = 3 / 6 = 0.5

# Probabilities always sum to 1
P(all outcomes) = 1

Joint Probability: Probability of two events both happening.

# P(A and B) = P(A) × P(B) if A and B are independent
# Example: probability of flipping heads AND rolling 6
P(heads and 6) = P(heads) × P(6) = 0.5 × (1/6) = 1/12

Conditional Probability: Probability of A given B already happened.

# P(A | B) = P(A and B) / P(B)

# Example: Hospital has 100 patients
# 10 have disease, 90 are healthy
# Test accuracy: 95% (true positive rate)

# If a patient tests positive, what's probability they actually have disease?
# P(disease | positive) = ?

# Let's use Bayes' theorem...

Part 2: Bayes' Theorem—Most Important Formula in AI

Formula: P(A|B) = P(B|A) × P(A) / P(B)

This is how doctors interpret test results, spam filters work, and ML models make predictions!

# Classic example: Disease testing
# Disease prevalence in population: 1% (P(disease) = 0.01)
# Test accuracy: 95% (correctly detects disease 95% of time)
# False positive rate: 10% (incorrectly says healthy person has disease)

P_disease = 0.01
P_healthy = 0.99
P_positive_if_disease = 0.95  # Test detects disease correctly
P_positive_if_healthy = 0.10  # False positive

# Someone tests positive. What's P(disease | positive)?
# First, find P(positive)
P_positive = P_positive_if_disease * P_disease + P_positive_if_healthy * P_healthy
P_positive = 0.95 * 0.01 + 0.10 * 0.99
P_positive = 0.0095 + 0.099 = 0.1085

# Apply Bayes
P_disease_given_positive = (P_positive_if_disease * P_disease) / P_positive
P_disease_given_positive = (0.95 * 0.01) / 0.1085
P_disease_given_positive = 0.0095 / 0.1085 ≈ 0.088 (8.8%)

# Surprising! Despite positive test, only 8.8% chance of disease!
# Because disease is rare, false positives dominate

print(f"P(disease | positive test) = {P_disease_given_positive:.1%}")  # 8.8%

This is why doctors often do follow-up tests—a single positive test is not conclusive when disease is rare!

Part 3: Probability Distributions

Discrete Distributions:

# Bernoulli: Binary outcome (0 or 1, heads or tails)
# P(X=1) = p, P(X=0) = 1-p

# Binomial: Number of successes in n trials
# Example: flip coin 5 times, how many heads?
from scipy.stats import binom

# n=5 trials, p=0.5 probability of heads
X = binom(n=5, p=0.5)
print(f"P(X=3) = {X.pmf(3):.3f}")  # P(exactly 3 heads) ≈ 0.3125

# Poisson: Number of events in time interval (rare events)
# Example: emails received per hour (average 10)
from scipy.stats import poisson

X = poisson(mu=10)
print(f"P(X=5) = {X.pmf(5):.4f}")  # P(exactly 5 emails) ≈ 0.0378

Continuous Distributions:

# Normal (Gaussian): Bell curve, very common
# μ = mean, σ = standard deviation

from scipy.stats import norm
import matplotlib.pyplot as plt

X = norm(loc=170, scale=10)  # μ=170cm, σ=10cm (human height)

# Probability of person being between 160-180cm
prob = X.cdf(180) - X.cdf(160)
print(f"P(160 < height < 180) = {prob:.2%}")  # ~95.4%

# Visualize
import numpy as np
x = np.linspace(140, 200, 1000)
y = X.pdf(x)

plt.plot(x, y)
plt.fill_between(x[(x >= 160) & (x <= 180)], y[(x >= 160) & (x <= 180)], alpha=0.3)
plt.xlabel('Height (cm)')
plt.ylabel('Probability Density')
plt.title('Normal Distribution of Heights')
plt.show()

# 68-95-99.7 rule (for any normal distribution)
print(f"P(μ-σ < X < μ+σ) = 68%")     # 160-180
print(f"P(μ-2σ < X < μ+2σ) = 95%")   # 150-190
print(f"P(μ-3σ < X < μ+3σ) = 99.7%") # 140-200

Exponential: Time between events (wait time, battery lifetime)

# λ = 1/mean_wait_time
# Example: average wait time at bus stop = 5 minutes
# What's probability of waiting < 2 minutes?

from scipy.stats import expon

X = expon(scale=5)  # scale = mean
prob = X.cdf(2)
print(f"P(wait < 2 min) = {prob:.2%}")  # ~33%

Part 4: Descriptive Statistics

import numpy as np

# Sample data: test scores of 10 students
scores = np.array([45, 52, 67, 78, 82, 88, 91, 73, 65, 79])

# Mean (average)
mean = np.mean(scores)  # 72.0

# Median (middle value when sorted)
median = np.median(scores)  # 75.5

# Mode (most frequent)
from scipy.stats import mode
mode_value = mode(scores).mode  # No repeats

# Variance (how spread out)
variance = np.var(scores)  # 217.6

# Standard deviation (square root of variance, same units as data)
std = np.std(scores)  # 14.75

print(f"Mean: {mean}, Median: {median}, Std: {std:.2f}")

# Percentiles
p25 = np.percentile(scores, 25)  # 25th percentile (Q1)
p50 = np.percentile(scores, 50)  # 50th percentile (median)
p75 = np.percentile(scores, 75)  # 75th percentile (Q3)

print(f"IQR (Interquartile range): {p75 - p25}")  # Q3 - Q1 = 20

Part 5: Hypothesis Testing

Question: A coin comes up heads 60 times out of 100 flips. Is it fair, or biased?

# Null hypothesis (H₀): Coin is fair (p = 0.5)
# Alternative hypothesis (H₁): Coin is biased (p ≠ 0.5)

from scipy.stats import binom_test

# Observed: 60 heads out of 100, expected 50 for fair coin
p_value = binom_test(60, n=100, p=0.5, alternative='two-sided')
print(f"p-value = {p_value:.4f}")

# If p-value < 0.05, reject null hypothesis (coin is biased)
# If p-value >= 0.05, fail to reject null hypothesis (could be fair)

# In this case: p ≈ 0.057, so we CANNOT conclusively say it's biased
# (Though it's close!)

Real Example: A/B Testing (Ubiquitous in Tech)

# Website test: Button color (red vs blue)
# Hypothesis: Blue button increases click rate

# Control (red): 500 users, 45 clicks → 9%
# Treatment (blue): 500 users, 65 clicks → 13%

# Question: Is this difference significant or just chance?

from scipy.stats import chi2_contingency

# Contingency table
table = np.array([
    [45, 455],   # Red button: clicks, no-clicks
    [65, 435]    # Blue button: clicks, no-clicks
])

chi2, p_value, dof, expected = chi2_contingency(table)
print(f"Chi-square = {chi2:.2f}, p-value = {p_value:.4f}")

# If p-value < 0.05: difference is significant, use blue button
# If p-value >= 0.05: difference might be chance, need more data

Part 6: Spam Classification with Bayes

Naive Bayes is one of the simplest yet powerful ML algorithms, directly from probability theory.

# Email classification: Spam or Not Spam?
# Features: presence of words like "free", "winner", "click here"

# Training data
spam_emails = [
    "Free money!! Click here!!!",
    "You won a prize! Claim now",
    "AMAZING OFFER - LIMITED TIME"
]

ham_emails = [
    "Meeting at 3 PM tomorrow",
    "Your order has been shipped",
    "See you this weekend"
]

# Vocabulary
words = set()
for email in spam_emails + ham_emails:
    words.update(email.lower().split())

# Count word frequencies
from collections import Counter

spam_word_counts = Counter()
for email in spam_emails:
    spam_word_counts.update(email.lower().split())

ham_word_counts = Counter()
for email in ham_emails:
    ham_word_counts.update(email.lower().split())

# For new email: "Click here for free money"
# P(spam | email) = P(email | spam) × P(spam) / P(email)

# P(email | spam) = P("click" | spam) × P("here" | spam) × ...
# This assumes word independence (naive!)

# Calculate probabilities
import numpy as np

email_to_classify = "click here for free".lower().split()

P_spam = len(spam_emails) / (len(spam_emails) + len(ham_emails))  # 0.5
P_ham = 1 - P_spam  # 0.5

P_email_given_spam = 1.0
for word in email_to_classify:
    if word in spam_word_counts:
        P_email_given_spam *= spam_word_counts[word] / sum(spam_word_counts.values())
    else:
        P_email_given_spam *= 1e-5  # Small probability for unseen words

P_email_given_ham = 1.0
for word in email_to_classify:
    if word in ham_word_counts:
        P_email_given_ham *= ham_word_counts[word] / sum(ham_word_counts.values())
    else:
        P_email_given_ham *= 1e-5

# Apply Bayes
numerator_spam = P_email_given_spam * P_spam
numerator_ham = P_email_given_ham * P_ham

P_spam_given_email = numerator_spam / (numerator_spam + numerator_ham)

print(f"P(spam | email) = {P_spam_given_email:.2%}")

if P_spam_given_email > 0.5:
    print("SPAM!")
else:
    print("Not spam")

Part 7: Confidence Intervals and Uncertainty Quantification

# We survey 1000 Indians, ask "Do you use smartphone?"
# 750 yes, 250 no
# Estimate: 75% use smartphones

# But we sampled only 1000 out of 1.4 billion. What's true population percentage?

# 95% Confidence interval
from scipy.stats import proportion_confint

proportion = 750 / 1000
n = 1000

lower, upper = proportion_confint(750, n, alpha=0.05, method='wilson')
print(f"95% CI: {lower:.2%} to {upper:.2%}")
# Might be: 71.9% to 78.1%

# Interpretation: We're 95% confident true population proportion is between 71.9% and 78.1%
# (NOT: 95% chance the true value is in this range—that's frequentist vs Bayesian subtlety)

Part 8: Bayes vs Frequentist

In practice: Frequentist for simple problems (hypothesis tests, A/B tests). Bayesian when you have prior knowledge (medical diagnosis, recommendation systems).

Part 9: Why This Matters for Your Career

Understanding probability and statistics is essential for:

• A/B testing: Every tech company tests new features. You need to know if improvement is real or chance.
• Model evaluation: What confidence interval on accuracy? Is model really 92% accurate or just lucky?
• Data analysis: Find real patterns vs noise
• Risk assessment: Finance, insurance, healthcare all use probability
• JEE/BITSAT: These exams heavily test probability and statistics

Deep Dive: Probability and Statistics for AI

At this level, we stop simplifying and start engaging with the real complexity of Probability and Statistics for AI. In production systems at companies like Flipkart, Razorpay, or Swiggy — all Indian companies processing millions of transactions daily — the concepts in this chapter are not academic exercises. They are engineering decisions that affect system reliability, user experience, and ultimately, business success.

The Indian tech ecosystem is at an inflection point. With initiatives like Digital India and India Stack (Aadhaar, UPI, DigiLocker), the country has built technology infrastructure that is genuinely world-leading. Understanding the technical foundations behind these systems — which is what this chapter covers — positions you to contribute to the next generation of Indian technology innovation.

Whether you are preparing for JEE, GATE, campus placements, or building your own products, the depth of understanding we develop here will serve you well. Let us go beyond surface-level knowledge.

Transformer Architecture: The Engine Behind GPT and Modern AI

The Transformer architecture, introduced in the landmark 2017 paper "Attention Is All You Need," revolutionised NLP and eventually all of deep learning. Here is the core mechanism:

# Self-Attention Mechanism (simplified)
import numpy as np

def self_attention(Q, K, V, d_k):
    """
    Q (Query): What am I looking for?
    K (Key):   What do I contain?
    V (Value): What do I actually provide?
    d_k:       Dimension of keys (for scaling)
    """
    # Step 1: Compute attention scores
    scores = np.matmul(Q, K.T) / np.sqrt(d_k)

    # Step 2: Softmax to get probabilities
    attention_weights = softmax(scores)

    # Step 3: Weighted sum of values
    output = np.matmul(attention_weights, V)
    return output

# Multi-Head Attention: Run multiple attention heads in parallel
# Each head learns different relationships:
# Head 1: syntactic relationships (subject-verb agreement)
# Head 2: semantic relationships (word meanings)
# Head 3: positional relationships (word order)
# Head 4: coreference (pronoun → noun it refers to)

The key insight of self-attention is that every token can attend to every other token simultaneously (unlike RNNs which process sequentially). This parallelism enables efficient GPU training. The computational complexity is O(n²·d) where n is sequence length and d is dimension, which is why context windows are a major engineering challenge.

State-of-the-art developments include: sparse attention (reducing O(n²) to O(n·√n)), mixture of experts (MoE — activating only a subset of parameters per input), retrieval-augmented generation (RAG — grounding responses in external documents), and constitutional AI (alignment through principles rather than RLHF alone). Indian researchers at institutions like IIT Bombay, IISc Bangalore, and Microsoft Research India are actively contributing to these frontiers.

India's Scale Challenges: Engineering for 1.4 Billion

Building technology for India presents unique engineering challenges that make it one of the most interesting markets in the world. UPI handles 10 billion transactions per month — more than all credit card transactions in the US combined. Aadhaar authenticates 100 million identities daily. Jio's network serves 400 million subscribers across 22 telecom circles. Hotstar streamed IPL to 50 million concurrent viewers — a world record. Each of these systems must handle India's diversity: 22 official languages, 28 states with different regulations, massive urban-rural connectivity gaps, and price-sensitive users expecting everything to work on ₹7,000 smartphones over patchy 4G connections. This is why Indian engineers are globally respected — if you can build systems that work in India, they will work anywhere.

Advanced Algorithms: Dynamic Programming and Graph Theory

Dynamic Programming (DP) solves complex problems by breaking them into overlapping subproblems. This is a favourite in competitive programming and interviews:

# Longest Common Subsequence — classic DP problem
# Used in: diff tools, DNA sequence alignment, version control

def lcs(s1, s2):
    m, n = len(s1), len(s2)
    dp = [[0] * (n + 1) for _ in range(m + 1)]

    for i in range(1, m + 1):
        for j in range(1, n + 1):
            if s1[i-1] == s2[j-1]:
                dp[i][j] = dp[i-1][j-1] + 1
            else:
                dp[i][j] = max(dp[i-1][j], dp[i][j-1])

    return dp[m][n]

# Dijkstra's Shortest Path — used by Google Maps!
import heapq

def dijkstra(graph, start):
    dist = {node: float('inf') for node in graph}
    dist[start] = 0
    pq = [(0, start)]  # (distance, node)

    while pq:
        d, u = heapq.heappop(pq)
        if d > dist[u]:
            continue
        for v, weight in graph[u]:
            if dist[u] + weight < dist[v]:
                dist[v] = dist[u] + weight
                heapq.heappush(pq, (dist[v], v))

    return dist

# Real use: Google Maps finding shortest route from
# Connaught Place to India Gate, considering traffic weights

Dijkstra's algorithm is how mapping applications find optimal routes. When you ask Google Maps to navigate from Mumbai to Pune, it models the road network as a weighted graph (intersections are nodes, roads are edges, travel time is weight) and runs a variant of Dijkstra's algorithm. Indian highways, city roads, and even railway networks can all be modelled this way. IRCTC's route optimisation for trains across 13,000+ stations uses graph algorithms at its core.

Research Frontiers and Open Problems in Probability and Statistics for AI

Beyond production engineering, probability and statistics for ai connects to active research frontiers where fundamental questions remain open. These are problems where your generation of computer scientists will make breakthroughs.

Quantum computing threatens to upend many of our assumptions. Shor's algorithm can factor large numbers efficiently on a quantum computer, which would break RSA encryption — the foundation of internet security. Post-quantum cryptography is an active research area, with NIST standardising new algorithms (CRYSTALS-Kyber, CRYSTALS-Dilithium) that resist quantum attacks. Indian researchers at IISER, IISc, and TIFR are contributing to both quantum computing hardware and post-quantum cryptographic algorithms.

AI safety and alignment is another frontier with direct connections to probability and statistics for ai. As AI systems become more capable, ensuring they behave as intended becomes critical. This involves formal verification (mathematically proving system properties), interpretability (understanding WHY a model makes certain decisions), and robustness (ensuring models do not fail catastrophically on edge cases). The Alignment Research Center and organisations like Anthropic are working on these problems, and Indian researchers are increasingly contributing.

Edge computing and the Internet of Things present new challenges: billions of devices with limited compute and connectivity. India's smart city initiatives and agricultural IoT deployments (soil sensors, weather stations, drone imaging) require algorithms that work with intermittent connectivity, limited battery, and constrained memory. This is fundamentally different from cloud computing and requires rethinking many assumptions.

Finally, the ethical dimensions: facial recognition in public spaces (deployed in several Indian cities), algorithmic bias in loan approvals and hiring, deepfakes in political campaigns, and data sovereignty questions about where Indian citizens' data should be stored. These are not just technical problems — they require CS expertise combined with ethics, law, and social science. The best engineers of the future will be those who understand both the technical implementation AND the societal implications. Your study of probability and statistics for ai is one step on that path.

Key Vocabulary

Here are important terms from this chapter that you should know:

🏗️ Architecture Challenge

Design the backend for India's election results system. Requirements: 10 lakh (1 million) polling booths reporting simultaneously, results must be accurate (no double-counting), real-time aggregation at constituency and state levels, public dashboard handling 100 million concurrent users, and complete audit trail. Consider: How do you ensure exactly-once delivery of results? (idempotency keys) How do you aggregate in real-time? (stream processing with Apache Flink) How do you serve 100M users? (CDN + read replicas + edge computing) How do you prevent tampering? (digital signatures + blockchain audit log) This is the kind of system design problem that separates senior engineers from staff engineers.

The Frontier

You now have a deep understanding of probability and statistics for ai — deep enough to apply it in production systems, discuss tradeoffs in system design interviews, and build upon it for research or entrepreneurship. But technology never stands still. The concepts in this chapter will evolve: quantum computing may change our assumptions about complexity, new architectures may replace current paradigms, and AI may automate parts of what engineers do today.

What will NOT change is the ability to think clearly about complex systems, to reason about tradeoffs, to learn quickly and adapt. These meta-skills are what truly matter. India's position in global technology is only growing stronger — from the India Stack to ISRO to the startup ecosystem to open-source contributions. You are part of this story. What you build next is up to you.

Crafted for Class 10–12 • Mathematics & Data Science • Aligned with NEP 2020 & CBSE Curriculum