NumPy Special Arrays

NumPy Special Arrays Explained

This guide focuses on how and why we use NumPy array creation functions — not just syntax.

---

🔹 Step 1: Import NumPy

import numpy as np

Important: We use np as a shortcut because NumPy functions are used frequently.

---

🔹 1. Zero Array → np.zeros()

ar_zero = np.zeros((3))
print(ar_zero)
print(ar_zero.ndim)

Output:

[0. 0. 0.]
1

✅ Explanation:

• (3) means 3 elements → 1D array
• All values are automatically set to 0
• .ndim tells dimension (1 here)

💡 Real Scenario:

Suppose you are building a system to track daily sales for 3 days. Initially, no sales → start with 0 values.

⚠️ Important Points:

• Always pass shape inside tuple
• (3) → 1D, (3,4) → 2D

---

🔸 Multi-Dimensional Example

ar_zero = np.zeros((3,4))
print(ar_zero)

Explanation:

• 3 rows and 4 columns
• Used when working with tables or grids

💡 Real Scenario:

Think of a classroom attendance sheet → 3 students, 4 subjects → all values start at 0.

---

🔹 2. Ones Array → np.ones()

ar_ones = np.ones((3))
print(ar_ones)

Output:

[1. 1. 1.]

✅ Explanation:

• Creates array filled with 1
• Structure same as zeros

💡 Real Scenario:

Used when all items have default value = 1 Example: Assign equal weight to 3 features in ML

⚠️ Important:

• Data type is float by default → 1. not 1

---

🔹 3. Empty Array → np.empty()

ar_empt = np.empty((3))
print(ar_empt)

✅ Explanation:

• Does NOT initialize values
• Returns whatever is in memory

💡 Real Scenario:

When you need speed and will overwrite values immediately.

⚠️ Important Concept:

If you previously created a ones array, memory may still contain [1. 1. 1.]. So empty array may show same values → but it is NOT actual data.

🔥 Hack:

• np.empty() is faster than np.zeros()
• Use only when you're sure you'll assign values

---

🔹 4. Range Array → np.arange()

ar_range = np.arange(3,7)
print(ar_range)

Output:

[3 4 5 6]

✅ Explanation:

• Starts from 3 and ends at 6 (7 excluded)
• Works like Python range but returns array

💡 Real Scenario:

Generating IDs → Customer IDs from 1000 to 1010

⚠️ Important:

• End value is always excluded

---

🔹 5. Diagonal Matrix → np.eye()

ar_diag = np.eye(4,4)
print(ar_diag)

✅ Explanation:

• Diagonal elements = 1
• Rest = 0

💡 Real Scenario:

Used in Machine Learning & Linear Algebra (Identity Matrix)

⚠️ Important:

• Can create non-square matrices too

np.eye(3,4)

---

🔹 6. Linspace → np.linspace()

ar_linespace = np.linspace(1,10,num=5)
print(ar_linespace)

Output:

[ 1.    3.25  5.5   7.75 10.  ]

✅ Explanation:

• Divides range into equal parts
• Includes both start and end

💡 Real Scenario:

Used in plotting graphs → evenly spaced X-axis values

⚠️ Important Difference:

• arange() → step-based
• linspace() → count-based

---

🔥 Final Key Takeaways

• zeros → initialize values
• ones → default weight/value
• empty → fast but risky
• arange → sequences
• eye → identity matrix
• linspace → equal spacing

---

🚀 Pro Tips

• Always check dimension using .ndim
• Use tuple correctly → common beginner mistake
• Prefer linspace for graphs
• Use empty only for performance optimization

---

❌ Common Mistakes

• Thinking empty = zeros
• Forgetting end is excluded in arange
• Confusing shape format

---

🔚 Conclusion

These functions are essential tools for working with NumPy efficiently.

Once you master them, handling data becomes much easier ⚡