Deep Learning Architectures and Tools

This unit develops practical fluency with the core mathematics, architectures, and tooling of deep learning: environment setup, frameworks (TensorFlow/Keras), linear algebra for network computation, and the inner workings of CNNs and LSTMs. Students apply these concepts to hands‑on problems (image classification, object perception, sequence modeling, pose estimation) and consider data, deployment, and ethical implications in real-world domains such as medicine and traffic safety.

Learning Objectives

• Demonstrate installation, configuration, and use of a deep learning programming environment (Anaconda, TensorFlow, and Keras) and download and organize datasets and example code for reproducible experiments
• Apply scalar, vector, and matrix operations and appropriate variable/data types to implement and debug forward and backward computations in basic neural network models
• Analyze and compare convolutional and recurrent (LSTM) architectures by designing, training, and evaluating models for image classification, object detection/pose estimation, and sequence memorization tasks
• Evaluate the effects of dataset size, augmentation, and model complexity (including transfer learning and augmentation via Keras) on performance and discuss real-world applications and ethical considerations in domains such as healthcare and autonomous systems