Backtracking

Backtracking is a general algorithmic technique for solving problems incrementally by trying partial
solutions and then abandoning (“backtracking”) them if they fail to satisfy the constraints.

🧠 Core Idea

Backtracking builds a solution one piece at a time and removes those solutions that fail to satisfy
constraints at any point during construction.

🧩 Typical Use Cases

 Combinatorial problems:

o N-Queens

o Sudoku solver

o Graph coloring

o Subset sum

o Permutations / Combinations

o Word search

✅ General Backtracking Steps

1. Choose a starting point.

2. Choose an option.

3. Check if it leads to a solution:

o ✅ If yes, move forward.

o ❌ If no, backtrack and try another option.

4. Repeat until all possibilities are explored or the solution is found.

🧬 Template (Pseudocode)

python

CopyEdit

def backtrack(solution):

if is_complete(solution):

process(solution)
 return

for option in choices(solution):

if is_valid(option, solution):

[Link](option)

backtrack(solution)

[Link]() # backtrack

🛠 Python Example: N-Queens (4x4)

python

CopyEdit

def solve_n_queens(n):

def is_safe(row, col, diagonals, anti_diagonals, cols):

return col not in cols and (row - col) not in diagonals and (row + col) not in anti_diagonals

def backtrack(row, diagonals, anti_diagonals, cols, state):

if row == n:

[Link](["".join(r) for r in state])

return

for col in range(n):

if is_safe(row, col, diagonals, anti_diagonals, cols):

state[row][col] = "Q"

[Link](row - col)

anti_diagonals.add(row + col)

[Link](col)

backtrack(row + 1, diagonals, anti_diagonals, cols, state)

# backtrack

state[row][col] = "."

[Link](row - col)

anti_diagonals.remove(row + col)
 [Link](col)

result = []

empty_board = [["." for _ in range(n)] for _ in range(n)]

backtrack(0, set(), set(), set(), empty_board)

return result

# Example: Print all solutions for 4-Queens

for board in solve_n_queens(4):

for row in board:

print(row)

print()

⏱ Time Complexity

 Worst-case is usually exponential, but it's faster than brute-force due to pruning.

o N-Queens: ~O(N!) (with optimizations)

o Sudoku: ~O(9^(N*N)), but pruned significantly in practice

✅ Advantages

 Simple, elegant recursive structure

 Solves complex constraint-based problems

 Easily extendable with optimization (like branch and bound)

❌ Limitations

 Can be slow for large input without optimization

 Requires good pruning conditions to be efficient

Exhaustive Search

Exhaustive Search (also called Brute Force Search) is a straightforward algorithmic technique that
systematically explores all possible solutions to a problem and selects the one(s) that satisfy the
given conditions.
🔍 Core Idea

Try every possible combination or path, regardless of efficiency, and pick the correct or optimal
solution.

🧠 When to Use

 When the problem space is small or moderate

 When correctness is critical

 As a baseline to test optimized algorithms

✅ Characteristics

 Guaranteed to find a solution (if one exists)

 Simple to implement

 Usually very slow for large inputs due to exponential growth

🕒 Time Complexity

 Typically O(kⁿ) where:

o k = number of choices at each step

o n = number of steps

🛠 Example 1: Subset Sum (Exhaustive Search)

python

CopyEdit

def subset_sum(nums, target):

def dfs(index, total):

if index == len(nums):

return total == target

# Include or exclude current number

return dfs(index + 1, total + nums[index]) or dfs(index + 1, total)

return dfs(0, 0)
# Example

print(subset_sum([3, 2, 7, 1], 6)) # True (3+2+1)

🛠 Example 2: Travelling Salesman Problem (TSP - Brute Force)

python

CopyEdit

from itertools import permutations

def tsp_brute_force(graph):

n = len(graph)

min_path = float('inf')

for perm in permutations(range(1, n)):

current_cost = graph[0][perm[0]]

for i in range(len(perm) - 1):

current_cost += graph[perm[i]][perm[i+1]]

current_cost += graph[perm[-1]][0] # return to start

min_path = min(min_path, current_cost)

return min_path

# Example Graph

graph = [

[0, 10, 15, 20],

[10, 0, 35, 25],

[15, 35, 0, 30],

[20, 25, 30, 0]

print(tsp_brute_force(graph)) # Output: 80

✅ Pros

 Easy to write and debug

 Always finds the correct result

❌ Cons

 Highly inefficient for large input sizes

 Doesn't scale well

 Often impractical in real-world scenarios

💡 Optimizations

To improve exhaustive search:

 Pruning (like in Backtracking)

 Memoization to avoid recomputation

 Heuristics to guide the search (informed search)

DFS (Depth-First Search)

Depth-First Search (DFS) is a graph/tree traversal algorithm that explores as far as possible along
each branch before backtracking.

✅ Key Concepts

 DFS uses stack data structure (implicitly via recursion or explicitly via a stack).

 It goes deep into the graph/tree before exploring siblings.

 Useful for:

o Finding connected components

o Topological sorting

o Detecting cycles

o Solving puzzles (mazes, Sudoku)

o Backtracking problems (e.g. N-Queens)

🧠 DFS Variants

 Recursive DFS – elegant and concise

 Iterative DFS – uses a stack to simulate recursion

🔁 Recursive DFS (Graph)

CopyEdit

def dfs_recursive(graph, node, visited):

if node in visited:

return

print(node, end=' ')

[Link](node)

for neighbor in graph[node]:

dfs_recursive(graph, neighbor, visited)

# Example usage:

graph = {

'A': ['B', 'C'],

'B': ['D', 'E'],

'C': ['F'],

'D': [],

'E': ['F'],

'F': []

visited = set()

dfs_recursive(graph, 'A', visited)

➡️Output: A B D E F C

📦 Iterative DFS (Using Stack)

python

CopyEdit

def dfs_iterative(graph, start):

visited = set()

stack = [start]

while stack:

node = [Link]()
 if node not in visited:

print(node, end=' ')

[Link](node)

[Link](reversed(graph[node])) # reverse for correct order

🕒 Time and Space Complexity

Graph Type Time Complexity Space Complexity

Adjacency List O(V + E) O(V)

Adjacency Matrix O(V²) O(V)

Where V = number of vertices, E = number of edges

🌐 DFS Applications

1. Cycle Detection in Directed/Undirected Graph

2. Topological Sorting (DAG)

3. Solving Mazes / Game Maps

4. Connected Components

5. Strongly Connected Components (Kosaraju’s Algorithm)

6. Pathfinding (less optimal than BFS in unweighted graphs)

⚠️DFS vs BFS

Feature DFS BFS

Data Structure Stack (Recursion) Queue

Memory Usage Less in wide graphs Less in deep graphs

Finds Shortest Path ❌ Not guaranteed ✅ Yes (in unweighted graphs)

Use Case Example Topo sort, cycle detection Shortest path, level order

What Are Constraints in Algorithmic Problems?

Constraints are rules or limits that define:

1. The range of inputs allowed

They are essential in algorithm design to choose the right strategy (brute-force, greedy, DP, etc.) and
to avoid Time Limit Exceeded (TLE) or Memory Limit Exceeded (MLE).

🧠 Types of Constraints

📌 1. Input Size Constraints

 Guide the choice of algorithm based on time complexity.

Input Size Max Operations Allowed Suggested Time Complexity

≤ 10 ~10! = 3.6 million Brute-force / Backtracking

≤ 100 ~10⁶ O(N³) acceptable

≤ 1,000 ~10⁶ O(N²) or better

≤ 10⁵ ~10⁸ O(N log N) or better

≤ 10⁶ ~10⁸ O(N) or better

📌 2. Value Constraints

 Help optimize memory usage and indexing strategies.

Example: 1 ≤ A[i] ≤ 1000 → can use arrays for frequency counting.

📌 3. Time Constraints

 Usually set to 1–2 seconds.

 On most platforms: 1 second ≈ 10⁸ operations

📌 4. Memory Constraints

 Typical limits: 256MB to 1GB

 Use these to estimate space complexity:

o int = 4 bytes

o bool = 1 byte

o list of 10⁶ ints ≈ 4MB

✅ How Constraints Guide Solution Strategy

Constraint Strategy

N ≤ 10 Brute force, permutations

N ≤ 20 Bit masking, backtracking

N ≤ 100 DP with O(N²)

N ≤ 10⁵ Greedy, sorting, binary search

N ≤ 10⁶ Hashing, prefix sums, linear scan

Values small (≤ 1000) Frequency arrays, counting sort

🛠 Example

Problem:

Given n integers (1 ≤ n ≤ 10⁵), find a pair with sum k.

 Brute-force O(n²) → TLE ❌

 HashSet O(n) → Accepted ✅

📦 Checklist for Using Constraints Effectively

 ✅ Read input size & limits carefully

 ✅ Estimate feasible time/space complexity

 ✅ Choose algorithm that fits the scale

 ✅ Watch for hidden constraints (sorted, unique, etc.)