## Overview

This competition consists of two parts with equal weighting:

| Component | Weight | Evaluation Method |
|-----------|--------|-------------------|
| Part 1: Denoising | 50% | Kaggle leaderboard (automated) |
| Part 2: Translation | 50% | Manual evaluation of translations |


#### **Final Score = 0.5 × Part1_Score + 0.5 × Part2_Score**

---

## Part 1: Denoising Submission

### Requirements

- Denoise the provided `test-part1-noise.csv` file
- **Submit as:** `<TeamID>_denoised_test_part1.csv`
- Maintain the same shape and format as the input file
- This file will be loaded in Part 2 of the competition

### Expected Format

Your submission CSV must follow this structure:

| ID | Col1 | Col2 | Col3 | ... | Col20 |
|----|------|------|------|-----|-------|
| 1  | 0.1  | 0.2  | 0.3  | ... | 0.9   |
| 2  | 0.05 | 0.15 | 0.25 | ... | 0.85  |
| 3  | 0.12 | 0.22 | 0.32 | ... | 0.92  |

### Submission Limits

- **Up to 10 submissions per day** to the Kaggle leaderboard
- Results visible immediately on public leaderboard
- You will be given marks based off your stnading in the private leaderboard

---

## Part 2: Translation Submission

### Requirements

- Modify the provided boilerplate notebook
- **Submit as:** `<TeamID>_part2`
- Share privately with TAs via their Kaggle IDs

| Name     | Kaggle ID |
| -------- | --------- |
| Vyoman     | vyomanjain  |
| Samarth   | creativestuffwithsam |
| Princia   | princiadsouza |
| Sudhanshu    | sudhanshusingh18 |

### Submission Limits

- You can only submit one notebook that is to be shared with all the TAs from your campus
- Failure to share the notebook or multiple submissions will result in a **25% penalty**

---

## Frequently Asked Questions

### General Questions

**Q: Can I use pretrained models?**  
A: Yes for Part 1 (denoising). For Part 2, you must use the provided LSTM architecture and weights.

**Q: Can I modify the LSTM architecture?**  
A: No, you must use the provided architecture for Part 2.

**Q: What programming language should I use?**  
A: Python is strongly recommended. All provided scripts are in Python.

### Strategy Questions

**Q: What if my denoising results are poor?**  
A: You can still excel in Part 2 through excellent training strategies. Part 2 is designed to give everyone a fair chance regardless of Part 1 performance.

**Q: How are submissions evaluated?**  
A: Part 1 is automatically scored via Kaggle leaderboard. Part 2 is manually evaluated by TAs based on translation quality.

---

## Resources & Support

### Where to Get Help

- **Kaggle Discussion Forum** - For questions and clarifications
- **TA Mentor** - Contact your assigned TA for guidance
- **FAQ Section** - Check regularly for updates

### Reporting Issues

- Technical issues should be reported immediately via the discussion forum
- FAQs will be updated regularly based on common questions

---

## Quick Reference Checklist

### Before Submitting Part 1
- [ ] File named correctly: `<TeamID>_denoised_test_part1.csv`
- [ ] Same shape as input file
- [ ] All columns present (ID, Col1-Col20)
- [ ] No missing values

### Before Submitting Part 2
- [ ] Notebook named correctly: `<TeamID>_part2`
- [ ] Shared privately with all 6 TAs
- [ ] Uses provided LSTM architecture
- [ ] Code runs without errors
- [ ] Results are reproducible

---

**Good luck with the competition!**