Deep Learning 🧠💻📘

Welcome to my Deep Learning and TensorFlow repository! This repository is a culmination of my learning journey in deep learning and TensorFlow, organized into three major sections: TensorFlow Basics, Car Price Prediction, and Malaria Diagnosis. Through this journey, I have explored foundational concepts, hands-on projects, and advanced techniques in deep learning. Below is a detailed breakdown of the contents.

1. 🔧 TensorFlow Basics

This section covers the foundational concepts of TensorFlow, focusing on tensor operations and essential functionalities.

• Basics: Understanding tensors and their properties.
• Initialization: Techniques for initializing tensors.
• Indexing: Manipulating tensor elements using indexing and slicing.
• Math Operations: Performing arithmetic operations on tensors.
• Linear Algebra Operations: Matrix operations and their applications in TensorFlow.
• Common TensorFlow Functions: Frequently used functions and their applications.
• Ragged Tensors: Working with tensors of irregular shapes.
• Sparse Tensors: Representing and manipulating sparse data.
• String Tensors: Handling string data in TensorFlow.
• Variables: Creating and updating variables in TensorFlow.

2. 🚗 Car Price Prediction

In this project, I explored linear regression to predict car prices. The process includes:

• Data Preparation: Cleaning and organizing the dataset for analysis.
• Linear Regression Model: Building a regression model to predict car prices.
• Error Sanctioning: Identifying and managing model errors.
• Training and Optimization: Enhancing model accuracy through optimization techniques.
• Performance Measurement: Evaluating model performance using appropriate metrics.
• Validation and Testing: Testing the model on unseen data to ensure reliability.
• Corrective Measures: Addressing underfitting or overfitting issues to improve model performance.

3. 🦠 Malaria Diagnosis

This section demonstrates a comprehensive deep learning pipeline to diagnose malaria using Convolutional Neural Networks (CNNs). It includes the following subsections:

• Data Prepration
  • Data Loading: Importing the malaria dataset.
  • Data Visualization: Exploring the dataset visually to understand patterns.
• Data Preprocessing
  • Data Partitioning: Splitting data into training, validation, and testing sets.
  • Data Augmentation: Enhancing the dataset with transformations.
  • Mixup Data Augmentation: Implementing the Mixup technique.
  • CutMix Data Augmentation: Applying CutMix for enhanced model generalization.
  • Albumentations: Leveraging advanced augmentation techniques.
• Model Creation
  • Sequential API:
    • Techniques like Dropout, Regularization, and Augmentation.
  • Functional API:
    • Building flexible models with callable layers.
  • Model Subclassing:
    • Creating custom layers and architectures.
  • Callbacks:
    • Tools like CSVLogger, EarlyStopping, LearningRateScheduler, ModelCheckpoint, and Tensorboard for monitoring and improving training.
    • Tensorboard Integration:
      • Data Logging, Visualizing Model Graphs, Hyperparameter Tuning, Profiling, and Visualizations.
• Modern Convolutional Neural Networks:
  • Implementing architectures such as AlexNet, VGGNet, ResNet, MobileNet, and EfficientNet.
  • Coding ResNet from Scratch.
• Training CNN
  • Custom Loss Functions and Metrics:
    • Custom implementations for loss and metrics with and without parameters.
    • Defining custom classes for advanced use cases.
  • Visualizations:
    • Visualizing training metrics and results.
• Model Evaluation and Testing
  • ROC and Confusion Matrix:
    • Assessing model performance on test data.
• Saving and Loading Models
  • Efficient model persistence techniques.