Submissions are evaluated according to the Mean Average Precision @ 3 (MAP@3):

$$MAP@3 = \frac{1}{U} \sum_{u=1}^{U}  \sum_{k=1}^{min(n,3)} P(k)$$

where `U` is the number of scored audio files in the test data, `P(k)` is the precision at cutoff `k`, and `n` is the number predictions per audio file.

You can learn more about this metric works from [this kernel](https://www.kaggle.com/wendykan/map-k-demo), and from this [python code](https://github.com/benhamner/Metrics/blob/master/Python/ml_metrics/average_precision.py).

## Submission File

For each `fname` (audio file) in the test set, you may predict up to 3 audio classification labels. The file should contain a header and have the following format:

    fname,label 
    00063640.wav,Laughter Hi-Hat Flute
    0013a1db.wav,Laughter Hi-Hat Flute
    ...