Oct 08, 2026 by Offer

TikTok Interview Questions 2027: Find Peak Element (Array & Matrix)

For these TikTok interview questions, use binary search for both: in 1D, compare mid with its right neighbor — rising means the peak is right, else left. In 2D, binary-search columns: take the middle column's max, move toward the larger neighbor. The invariant: always keep the half guaranteed to hold a peak.

TikTok Interview Questions: What the Peak Problems Test

The peak-element pair is commonly reported by candidates in TikTok interviews because it tests whether you can extend binary search beyond sorted-array lookup. The 1D version checks that you understand binary search as "discard the half that cannot contain the answer." The 2D version tests whether you can lift that reasoning into two dimensions. Interviewers watch for the invariant: can you articulate why the remaining half must contain a peak?

How to Solve These TikTok Interview Questions Step by Step

1D peak element:

State the invariant. "I'll maintain that a peak exists within [left, right]. Comparing mid with mid+1 tells me which half to keep."
The decision. If nums[mid] < nums[mid+1], the sequence rises rightward, so a peak lies in (mid, right] — set left = mid + 1. Otherwise a peak lies in [left, mid] — set right = mid.
Terminate. When left == right, that index is a peak — the virtual -infinity beyond both ends guarantees one exists.
Complexity. O(log n) time, O(1) space.

2D peak element:

Reduce to 1D. "I'll binary-search over columns: for the middle column, find its maximum element."
Compare sideways. Let (r, c) be that max. If the left neighbor is larger, the peak lies in the left half — recurse there. If the right neighbor is larger, go right. Otherwise (r, c) is a 2D peak.
Justify the invariant. The column max beats everything in its column, so moving toward the larger neighbor keeps a guaranteed peak in the remaining half — the same rising argument as 1D.
Complexity. O(m log n) time for an m×n matrix (each step scans one column), O(1) extra space.

Example line: "The unifying idea is the rising-slope argument — wherever the terrain rises toward, a peak must lie beyond, so I can always discard the downhill half."

Common Mistakes

1D: comparing with both neighbors. You only need mid vs. mid+1 — the invariant handles the rest.
2D: picking any element in the column. It must be the column maximum, or the invariant breaks.
Off-by-one in bounds. Use the left < right loop form with right = mid (not mid - 1).

Keep Reading

FAQ

Why does a peak always exist? Treat out-of-bounds as negative infinity — then the global maximum is always a peak, so at least one exists.

Can I solve 1D in linear time? Yes, but the interviewer wants the O(log n) binary search — that is the point of the question.

Is the 2D approach the only one? A greedy hill-climbing approach also works but has worse worst-case complexity; column binary search is the expected answer.

What if the matrix has one row? It reduces exactly to the 1D problem — mention this to show you see the connection.

Preparing for TikTok's interview? Our 2027 TikTok Online Hackerrank Coding Assessment Tutorials has practice questions and answers — $79 one-time, instant download.