IDTS5102 - Deep Learning and Applications

This course provides a comprehensive understanding of the theoretical foundations and practical implementation of Deep Learning techniques, systematically mapping advanced neural network architectures to real-world deployment scenarios. The curriculum bridges basic mathematical formulations with industrial-grade predictive frameworks.

Students will explore the foundational components of deep models, starting with core linear algebra matrix structures, probability theory constraints, and simple perceptron separations. The training focuses on the mathematical mechanics of backpropagation, chain-rule optimization loops, gradient descent variants, and deep regularization methods (such as Dropout and Batch Normalization). The course covers advanced deep architectures, including Convolutional Neural Networks (CNN) for geospatial spatial matrices and computer vision layers, Recurrent Neural Networks (RNN) alongside Long Short-Term Memory (LSTM) gates for sequential token analysis, and generative transformer blocks.

Additionally, the course addresses critical infrastructure governance, profiling bias auditing, and personal data masking rules in accordance with regulatory laws (such as the Indonesian UU PDP and GDPR) within deep learning classifiers. Through extensive hands-on labs and project-based assignments, students will develop the professional ability to build, optimize, and defend scalable intelligent neural architectures.