Join Our 5-Week ML/AI Engineer Interview Bootcamp 🚀 led by ML Tech Leads at FAANGs
47. Sin and cosine feature encoding
GeneralSinusoidal encoding turns a cyclic feature (like hour-of-day) into two smooth features using sine and cosine so models don’t treat the “end” and “start” as far apart. You’ll implement this encoding for an array of cyclic values and return the encoded features as NumPy arrays.
Requirements
Implement the function
Rules:
- Use the formula above to compute one sine and one cosine value per input.
- Return a tuple of two NumPy arrays: all sine values, then all cosine values.
- Use NumPy for vectorized math (
np.sin,np.cos,np.pi). - Do not use any feature-engineering helper libraries (e.g., sklearn encoders).
Example
Output:
| Argument | Type |
|---|---|
| period | float |
| values | np.ndarray |
| Return Name | Type |
|---|---|
| value | tuple |
Constraints
Use NumPy sin/cos; return NumPy arrays.
No sklearn/feature-engineering helper libraries.
Output tuple of np.ndarrays: (sin_arr, cos_arr).
Hint 1
Map each value onto the unit circle by converting it to an angle in radians: angle = 2π * x / period.
Hint 2
Use NumPy element-wise operations to compute angle, sin(angle), and cos(angle) for the entire array at once.
Hint 3
Use np.pi for π, and ensure your input is treated as a NumPy array (which it is). Return np.sin(angles), np.cos(angles).
Roles
Companies
GeneralLevels
Tags
Python Editor