Correlation with np.correlate() in NumPy - Cheat Sheet & Quick Revision

Recall & Review

beginner

What does the function np.correlate() compute?

np.correlate() computes the cross-correlation of two 1-dimensional sequences. It measures how similar two sequences are as one is shifted over the other.

intermediate

What is the difference between correlation and convolution in numpy?

Correlation compares two sequences by sliding one over the other without flipping. Convolution flips one sequence before sliding. np.correlate() does correlation, while np.convolve() does convolution.

beginner

What does the mode parameter in np.correlate() control?

The mode parameter controls the size of the output:

• 'valid': Only positions where sequences fully overlap.
• 'same': Output is the same length as the first sequence.
• 'full': Includes all overlaps, even partial.

intermediate

How can np.correlate() be used to find a pattern in a signal?

By correlating the signal with the pattern, peaks in the result show where the pattern matches the signal best. This helps detect the pattern's position inside the signal.

beginner

What is the shape of the output when using np.correlate(a, v, mode='full') if a has length n and v has length m?

The output length will be n + m - 1. This includes all possible overlaps of the two sequences.

What does np.correlate(a, v, mode='valid') return?

A An output with the same length as <code>a</code>

B The full correlation including partial overlaps

C Only the parts where <code>a</code> and <code>v</code> completely overlap

D The convolution of <code>a</code> and <code>v</code>

mode='valid' returns correlation values only where the sequences fully overlap.

If a = [1, 2, 3] and v = [0, 1], what is the length of np.correlate(a, v, mode='full')?

A 4

B 3

C 2

D 5

Length is 3 + 2 - 1 = 4.

Which numpy function flips one sequence before sliding for comparison?

A np.correlate()

B np.convolve()

C np.dot()

D np.cross()

np.convolve() flips one sequence before sliding, unlike np.correlate().

What does a peak in the output of np.correlate(signal, pattern) indicate?

A A strong similarity or match at that position

B A mismatch between signal and pattern

C The length of the pattern

D The sum of signal values

Peaks show where the pattern matches the signal well.

Which mode in np.correlate() gives output length equal to the first input's length?

A valid

B none

C full

D same

mode='same' returns output with the same length as the first input.

Explain in your own words what np.correlate() does and how it can be used in real life.

Think about how you might find a small sound or pattern inside a longer recording.

Describe the effect of different mode options in np.correlate() and when you might use each.

Consider if you want only exact matches or partial matches included.