Deep Learning Principles & Applications

Home
Courses
Deep Learning Principles & Applications

Generative AI is at the forefront of technological innovations, with applications across diverse industries like healthcare, retail, sales, legal, and entertainment. At the core of this revolution are deep learning techniques that power large language models (LLMs), enabling the development of sophisticated AI-driven applications.

This 10-day workshop provides a hands-on, computation-focused introduction to deep learning for language modeling and Generative AI. Participants will gain essential skills to build and apply cutting-edge LLMs in real-world scenarios. The workshop balances foundational theory with practical implementation, offering a streamlined yet accessible introduction to key concepts from linear algebra, deep learning, and natural language processing (NLP) through interactive coding sessions. Participants will explore the inner workings of LLMs, experiment with state-of-the-art AI frameworks, and develop innovative solutions in NLP. By the end of the workshop, attendees will be equipped with the knowledge and tools to harness Generative AI for impactful applications across various domains.

Unit name	Manipal School of Information Sciences
Month, Year	July 2025
Duration	Two Weeks
Attendance mode	Regular
Location	Manipal Academy of Higher Education, MAHE
ECTS	3

Day	Topic	Learning Outcome
Day 1	Introduction to essential linear algebra for deep learning using PyTorch.	LO1
Day 2	Application project: analyzing words in Wikipedia articles using static embeddings.	LO1
Day 3	The softmax classifier for linear classification and implementation using model subclassing in PyTorch.	LO1
Day 4	Application project: sentiment analysis of product reviews.	LO2
Day 5	Foundations of deep neural networks and implementation using model subclassing in PyTorch.	LO2
Saturday	Application project: sentiment analysis of product reviews.
Sunday	Vector semantics, word embeddings, and preprocessing raw textual data.	LO3
Day 6	Application project: preprocessing and analyzing the Reuters corpus.	LO3
Day 7	Learning word embeddings using the Word2Vec algorithm.	LO4
Day 8	Application project: embedding and visualizing words from the Reuters corpus.	LO4
Day 9	Weekend – KAIROS 2025
Day 10	The self-attention mechanism in language modeling and the transformer neural network
Saturday	Application project: analyzing static embeddings of words from Wikipedia articles and extending them to contextual embeddings.
Sunday	Transformer architectures for language modeling using PyTorch and Hugging Face.
Day 11	Application project: load models and their inferences and train models with Hugging Face.
Day 12	Large language models with transformers.
Day 13	Application project: sentence classification using the BookCorpus dataset using BERT.
Day 14	Pre-training and fine tuning large language models.
Day 15	Application project: build a movie recommendation chatbot.
Day 16	Final quiz and review day.
Day 17	Closure & Departure

Type	Description	Weightage	Date	Mode
Class participation	Attendance and active participation in class meetings	50%	Day 1-10	In-class
Final Quiz	Multiple choice quiz for 30 minutes duration on Day 1-7 topics	50%	On Day-10	In-lab, online submission

Learning Outcomes

Course Content

Essential linear algebra for deep learning
Fundamentals of linear classification: weights, bias, scores, and loss functions
Calculus for the gradient descent algorithm
Forward and backward propagation with regularization
Batch processing for large datasets
Linear to nonlinear classification via activation functions
Computational setup of a shallow neural network
Tuning neural network performance
Pre-processing data and batch normalization
Cross-validation for validating model performance
Extending the computational setup from a shallow to a deep neural network
Introduction to the TensorFlow library
Application projects: implementing shallow and deep neural network models using TensorFlow; implementing machine learning models on edge devices using Edge Impulse.