Greedy Algorithm Technique in DSA

What is the Greedy Algorithm Technique?

The Greedy Algorithm Technique is a method for solving optimization problems by making a series of choices, each of which looks best at the moment. Unlike dynamic programming, greedy algorithms do not revisit or revise earlier decisions.

This method is effective when the problem has the greedy-choice property (local optimum leads to global optimum) and optimal substructure (optimal solution of a problem contains optimal solutions to subproblems).

What is Greedy-Choice Property?

The greedy-choice property refers to a situation where making the best possible choice at each small step leads to the best overall solution. That is, you don’t need to look ahead or try all possible combinations — you can build the solution step-by-step by always picking what looks best right now.

This works only if every local decision (greedy choice) contributes to the global optimum. If even one local choice can lead you away from the best final answer, then the problem does not satisfy the greedy-choice property. So, a problem that has this property lets you be short-sighted and still guarantees that you’ll get the best answer in the end.

Intuitively, imagine you're trying to reach a goal by choosing one option at a time. If always taking the seemingly best step gets you to the best final result — without ever needing to go back and change your decisions — then the greedy-choice property holds.

If a problem does not have this property, greedy strategies may lead to incorrect or sub-optimal results.

What is Optimal Substructure?

Optimal substructure means that a big problem can be broken down into smaller pieces, and the best solution to the big problem depends on the best solutions to those smaller pieces.

This property is crucial because it allows you to solve the overall problem by solving its parts one at a time. If the subproblems are solved optimally, then combining them gives the correct solution for the full problem.

To build intuition: imagine you're solving a puzzle where solving smaller sections perfectly leads to the whole puzzle being solved. If fixing the small parts automatically helps fix the entire picture, then the puzzle has optimal substructure.

In short, optimal substructure allows you to build the global solution incrementally by solving smaller parts optimally.

Examples of Greedy Algorithm Applications

1. Coin Change Problem (Minimum Coins)

Problem Statement: You are given a set of coin denominations (like 1, 2, 5, 10, etc.) and a target amount. The goal is to make the target amount using the fewest number of coins. You are allowed to use an unlimited number of each coin type.

Example

Suppose the available coins are [1, 5, 10, 20, 50, 100] and the target amount is 135.

Greedy Strategy

The greedy approach says: Always choose the largest coin that does not exceed the remaining amount. Subtract that coin from the amount, and repeat the process until the amount becomes zero.

Step-by-Step Breakdown:

• Start with amount = 135
• The largest coin ≤ 135 is 100 → use it → remaining = 35
• The largest coin ≤ 35 is 20 → use it → remaining = 15
• The largest coin ≤ 15 is 10 → use it → remaining = 5
• The largest coin ≤ 5 is 5 → use it → remaining = 0

Coins used: [100, 20, 10, 5] → total coins = 4

Pseudocode

// Pseudocode (assuming denominations are sorted in descending order)
count = 0
for coin in coins:
    while amount >= coin:
        amount -= coin
        count += 1
return count

Why Greedy Works Here

Greedy-Choice Property:

This problem has the greedy-choice property if the coin system is designed such that choosing the largest denomination at each step always leads to the optimal solution. In standard currency systems (like Indian Rupees or US Dollars), this property is satisfied. Taking the largest coin first always minimizes the total number of coins.

Why? Because skipping a larger coin in favor of multiple smaller coins would always increase the total number of coins used, making it worse.

Optimal Substructure:

The optimal solution to the overall problem depends on the optimal solution to smaller subproblems.

For example, to solve the problem for amount = 135:

• We first take 100, and now we need to solve for 35.
• Then we take 20, and solve for 15.
• Then take 10, and solve for 5.
• Then take 5, and we’re done.

Each of these steps is a smaller version of the same problem (finding the best way to make a smaller amount), and solving each optimally helps solve the full problem optimally.

Time and Space Complexity

• Time Complexity: O(n) in the best case (if we quickly reach the amount), or O(n * amount / smallest coin) in worst case when using many small coins.
• Space Complexity: O(1) — no extra memory used other than counters.

Important Note:

This greedy solution does not work with all coin systems. For example, if coins are [1, 3, 4] and the target is 6, the greedy approach gives [4, 1, 1] (3 coins), but the optimal is [3, 3] (2 coins). That’s why the coin system must be canonical — designed so that the greedy algorithm always gives the correct answer.

How to Solve Such Cases where Greedy Algorithm Does not Work?

• Dynamic Programming: Try all combinations in a bottom-up or top-down approach and store solutions for subproblems.
• Backtracking or BFS: Explore all valid combinations (not scalable for large inputs).

The coin change problem is an excellent example of how greedy algorithms can solve optimization problems efficiently — but only when the problem satisfies both the greedy-choice property and optimal substructure. It's crucial to test whether the greedy method works for your specific input set before applying it in production.

2. Activity Selection Problem

Problem Statement:

You are given a list of activities. Each activity has a start time and an end time. Your goal is to select the maximum number of non-overlapping activities — meaning no two selected activities can happen at the same time.

This is a classic optimization problem where you must choose activities in such a way that you can attend the maximum number without any time conflicts.

Understanding the Intuition

Imagine you are organizing a schedule. You have multiple tasks to complete, but you can only do one at a time. Each task takes up a block of time. Your aim is to pick the highest number of tasks without overlaps, so that your time is used efficiently.

At first glance, you might think of trying all combinations of activities and choosing the one with the most tasks. But that would be slow (exponential time). Instead, we use a greedy algorithm.

Why Greedy Works Here

Greedy-Choice Property

We claim that if we always choose the activity that finishes the earliest (among the remaining ones), we will eventually arrive at the best solution. This is the greedy choice. The reason it works is because:

• By finishing early, we leave the most room for future activities.
• Choosing an activity that finishes late may block multiple future opportunities.

Optimal Substructure

Once we pick an activity (say, the one that ends earliest), the problem reduces to a smaller subproblem: selecting the maximum number of activities that start after the one we just chose. If we solve this subproblem optimally and combine it with our first greedy choice, we get an optimal solution for the entire problem.

Example

Consider the following list of activities:

Activity	Start	End
A1	1	4
A2	3	5
A3	0	6
A4	5	7
A5	8	9
A6	5	9
A7	6	10

Step 1: Sort Activities by End Time

We sort the activities based on their end times:

Activity	Start	End
A1	1	4
A2	3	5
A3	0	6
A4	5	7
A5	8	9
A6	5	9
A7	6	10

Step 2: Greedy Selection

We start by picking A1 (ends at 4).

Next, we look for the next activity whose start time is ≥ 4. A2 starts at 3, so skip. A3 starts at 0, skip. A4 starts at 5 → select A4.

Now A4 ends at 7. Next activity starting ≥ 7 is A5 (starts at 8) → select A5.

Final selected activities: A1, A4, A5

Maximum number of non-overlapping activities = 3

Pseudocode

# Let activities be a list of (start, end) tuples
activities.sort(key=lambda x: x[1])  # Sort by end time
last_end = -1
selected = []

for activity in activities:
    if activity[0] >= last_end:
        selected.append(activity)
        last_end = activity[1]

return selected

Time Complexity:

• O(n log n) — for sorting the activities by end time.
• O(n) — to iterate and select activities.

Space Complexity:

• O(1) — if we just count them.
• O(k) — if we store selected activities (where k is number of selected activities).

The Activity Selection Problem is a great example of a greedy algorithm that works efficiently and correctly due to two key properties:

• Greedy-Choice Property: Always picking the earliest finishing activity leads to the optimal result.
• Optimal Substructure: The remaining problem after selecting one activity is a smaller version of the same problem.

Because of these properties, the greedy method provides an optimal and efficient solution without backtracking or complex logic.

// Pseudocode using a priority queue (min-heap)
while (more than one node in queue):
    node1 = extract min frequency node
    node2 = extract next min frequency node
    mergedNode = new node with freq = node1.freq + node2.freq
    insert mergedNode back into queue
return root of Huffman Tree

4. Fractional Knapsack Problem — Explained for Beginners

Problem Statement:

You are given a set of items. Each item has a value (profit) and a weight. You also have a knapsack (bag) that can hold a maximum weight W.

Your goal is to maximize the total value you can carry in the knapsack. However, there's a twist — you are allowed to take fractions of items. So if an item weighs 10 units and you only have 5 units of capacity left, you can take half of it and get half the value.

Greedy Strategy:

To get the most value for the least weight, we follow a greedy approach. For each item, we compute its value-to-weight ratio. This tells us how much value we get per unit of weight. We then:

1. Sort all items in descending order of their value-to-weight ratio.
2. Pick the item with the highest ratio first and take as much of it as we can.
3. If we can’t take the whole item, we take as much as the remaining capacity allows (a fraction).
4. Repeat until the knapsack is full.

Example:

Suppose you have the following items and knapsack capacity W = 50:

Item	Value	Weight
1	60	10
2	100	20
3	120	30

Step 1: Calculate value/weight ratio

• Item 1: 60/10 = 6
• Item 2: 100/20 = 5
• Item 3: 120/30 = 4

Step 2: Sort items by value/weight ratio: Item 1 (6), Item 2 (5), Item 3 (4)

Step 3: Start filling the knapsack (capacity = 50)

• Take all of Item 1 (weight = 10, value = 60). Remaining capacity = 40.
• Take all of Item 2 (weight = 20, value = 100). Remaining capacity = 20.
• Take 2/3 of Item 3 (weight = 20, value = (2/3)*120 = 80). Remaining capacity = 0.

Total value = 60 + 100 + 80 = 240

How This Satisfies Greedy-Choice Property:

At every step, we picked the item that gave us the most value per unit weight — the locally optimal (greedy) choice. Because the problem allows breaking items and the value scales linearly with weight, this strategy always leads to the optimal (best) global solution. We never had to go back and change earlier decisions.

How This Satisfies Optimal Substructure:

Suppose we’ve filled part of the knapsack optimally (say the first 30 units). The remaining 20 units of capacity is now a smaller subproblem. Solving it using the same strategy and combining both gives us the best answer for the entire problem. So, solving smaller parts optimally builds the global solution — this is optimal substructure.

Pseudocode:

function fractionalKnapsack(items, W):
    sort items by value/weight in descending order
    totalValue = 0
    for item in items:
        if item.weight <= W:
            totalValue += item.value
            W -= item.weight
        else:
            totalValue += item.value * (W / item.weight)
            break  // Knapsack is full
    return totalValue

Time and Space Complexity:

• Time Complexity: O(n log n) — due to sorting the items by ratio.
• Space Complexity: O(1) — if sorting is done in-place, and no extra data structures are used.

Conclusion: Fractional Knapsack is a classic example of a problem where the greedy algorithm guarantees the optimal solution. This is because each choice is locally best and contributes perfectly to the global solution due to the linear scaling of value with weight.

Advantages and Disadvantages of Greedy Technique

Advantages

• Fast and Efficient: Greedy algorithms are generally faster than dynamic programming or brute force.
• Simpler to Implement: They often require fewer lines of code and simpler logic.
• Scalable: Performs well even for large datasets if conditions are met.

Disadvantages

• Not Always Optimal: May fail to produce correct results if greedy-choice property is not satisfied.
• Problem-Specific: Requires deep understanding of problem structure.
• No Backtracking: Once a decision is made, it’s not revised.

Conclusion

Greedy algorithms are powerful tools when used in the right problems. They help build fast and efficient solutions using local decisions. However, it’s essential to validate that the problem meets the greedy criteria. Otherwise, the solution might be sub-optimal or incorrect. Use them wisely after testing against edge cases or comparing with optimal solutions.