About This Hackathon

Speed ML Challenge: Handwritten Digit Classification @ MLDS 2025Get ready for an exclusive 1-hour challenge at MLDS 2025 that will push your deep learning and model optimization skills to the limit! This fast-paced event is designed to test how efficiently you can train and optimize a neural network model for classifying handwritten digits.<h3>Event Details</h3>What You Need to Do: Develop a robust neural network model that accurately classifies handwritten digits while minimizing inference time.A test dataset will be released during the competition, and participants must generate predictions within the given timeframe.Your submission will be evaluated based on a 50-50 weighted average of accuracy and inference time.This challenge is an exciting addition to MLDS 2025 giving participants a chance to showcase their deep learning prowess in a rapid competition format!<h3>Participation and Benefits</h3>✅ Beginner-Friendly: Ideal for anyone familiar with deep learning and eager to compete in a high-stakes, time-sensitive challenge.  ✅ Certificates & Recognition: Every participant will receive a certificate from MachineHack, and winners will secure a place on the leaderboard.<h3>Submission and Evaluation</h3>📌 Submission Format: Participants must submit their predictions in the format specified in submission.csv.  📌 Evaluation Metric: Submissions will be ranked based on a 50-50 weighted accuracy and inference time average.  📌 Leaderboard: Track your progress in real-time and strive for the top position!<h3>Dataset Description</h3>The dataset includes:  📂 Test.zip & Test.csv (Will be available during the competition itself) – Contains images and image labels respectively for which predictions must be generated.  📂 StarterNotebook.pynb – A starter notebook which you have to work-out and improve to get the better and better accuracy. 📂 Train.zip – A zip file containing all the images required for training and testing.  📂 Submission.csv – Template format for submitting predictions.Note: Please submit predictions in the submission file in the ascending order of the file names.<h3>How to Approach the Challenge</h3>🔹 Data Preprocessing: Normalize pixel values for better model convergence. Apply data augmentation (rotation, flipping, scaling) to enhance model robustness.🔹 Model Development: Use CNN architectures like LeNet, ResNet, or EfficientNet. Experiment with custom models to optimize performance.🔹 Training & Optimization: Implement learning rate scheduling, dropout, and batch normalization to boost generalization. Utilize TensorFlow, PyTorch, or Keras for efficient training.🔹 Inference Speed Optimization: Optimize model architecture to minimize prediction time. Quantization and model pruning can be leveraged for faster inference.Getting Started: 🔹 Register Now: Secure your spot in this exclusive MLDS 2024 challenge.  🔹 Download the Dataset: The dataset will be provided on the MachineHack platform.<h3>Support & Resources</h3>For any queries, reach out to support@machinehack.com. Stay updated with announcements by subscribing to our newsletter.🚀 Happy Hacking & Best of Luck!

Speed ML Challenge: Handwritten Digit Classification @ MLDS 2025

Get ready for an exclusive 1-hour challenge at MLDS 2025 that will push your deep learning and model optimization skills to the limit! This fast-paced event is designed to test how efficiently you can train and optimize a neural network model for classifying handwritten digits.

Event Details

What You Need to Do: Develop a robust neural network model that accurately classifies handwritten digits while minimizing inference time.

A test dataset will be released during the competition, and participants must generate predictions within the given timeframe.

Your submission will be evaluated based on a 50-50 weighted average of accuracy and inference time.

This challenge is an exciting addition to MLDS 2025 giving participants a chance to showcase their deep learning prowess in a rapid competition format!

Participation and Benefits

✅ Beginner-Friendly: Ideal for anyone familiar with deep learning and eager to compete in a high-stakes, time-sensitive challenge.
✅ Certificates & Recognition: Every participant will receive a certificate from MachineHack, and winners will secure a place on the leaderboard.

Submission and Evaluation

📌 Submission Format: Participants must submit their predictions in the format specified in submission.csv.
📌 Evaluation Metric: Submissions will be ranked based on a 50-50 weighted accuracy and inference time average.
📌 Leaderboard: Track your progress in real-time and strive for the top position!

Dataset Description

The dataset includes:
📂 Test.zip & Test.csv (Will be available during the competition itself) – Contains images and image labels respectively for which predictions must be generated.
📂 StarterNotebook.pynb – A starter notebook which you have to work-out and improve to get the better and better accuracy.
📂 Train.zip – A zip file containing all the images required for training and testing.
📂 Submission.csv – Template format for submitting predictions.

Note: Please submit predictions in the submission file in the ascending order of the file names.

How to Approach the Challenge

🔹 Data Preprocessing:
Normalize pixel values for better model convergence.
Apply data augmentation (rotation, flipping, scaling) to enhance model robustness.

🔹 Model Development:
Use CNN architectures like LeNet, ResNet, or EfficientNet.
Experiment with custom models to optimize performance.

🔹 Training & Optimization:
Implement learning rate scheduling, dropout, and batch normalization to boost generalization.
Utilize TensorFlow, PyTorch, or Keras for efficient training.

🔹 Inference Speed Optimization:
Optimize model architecture to minimize prediction time.
Quantization and model pruning can be leveraged for faster inference.

Getting Started:
🔹 Register Now: Secure your spot in this exclusive MLDS 2024 challenge.
🔹 Download the Dataset: The dataset will be provided on the MachineHack platform.

Support & Resources

For any queries, reach out to support@machinehack.com. Stay updated with announcements by subscribing to our newsletter.