Deep Learning Approaches Comparison on the Diabetes Dataset

This project compares two different deep learning approaches using the Diabetes dataset:

1. Using Keras & TensorFlow with a high-level API
2. Without any deep learning frameworks (built from scratch using Python & NumPy)

🎯 Objective

To observe the differences between high-level frameworks and manual development in terms of speed, optimization, and flexibility, as well as the learning depth gained from implementing neural networks from scratch.

🛠 Technologies Used

Without Frameworks (From Scratch)

• Python
• NumPy
• Pandas
• Matplotlib
• scikit-learn (for preprocessing and metrics)

Keras & TensorFlow Model

• Python
• TensorFlow / Keras
• Pandas
• NumPy
• scikit-learn

📊 Summary of Comparison

Feature	Keras & TensorFlow	From Scratch
Development Speed	Very fast	Slower
Optimization	GPU support, automatic	CPU only, manual
Learning Depth	Moderate	High
Batch Normalization / Dropout	Yes	Not implemented (manual possible)