Submissions are evaluated based on the [Area Under the ROC Curve (AUC)](https://en.wikipedia.org/wiki/Receiver_operating_characteristic) between the predicted probability and the actual `is_dog` target.

Your goal is to predict how likely each image contains a **dog**.
Higher AUC scores indicate better model performance at distinguishing between cats and dogs.

---

### **Submission File**

For each image in the dataset, you must predict the probability that it is a dog (`is_dog = 1`).
Your submission file should include a header and follow this format:

```
img,is_dog
0,0
1,1
2,1
3,0
...
```

Where:

* `img` — corresponds to the image filename without extension (e.g., `0` for `0.jpg`)
* `is_dog` — is the **predicted probability** (a float between 0 and 1) that the image is a dog.