Dr. Jerrin Thomas Panachakel and Prof. Jithuna T.S.
https://meet.google.com/pmp-dybr-mzb
Module 1: Basics of Machine Learning (8 hours)
Basics of machine learning, supervised and unsupervised learning, examples
Features, feature vector, training set, target vector, test set
Over-fitting, curse of dimensionality
Evaluation and model selection: ROC curves, evaluation measures
Validation set, bias-variance trade-off
Confusion matrix, recall, precision, accuracy
Module 2: Regression and Classification (7 hours)
Regression: linear regression, error functions in regression
Multivariate regression, regression applications, bias and variance
Classification: Bayes’ decision theory
Discriminant functions and decision surfaces
Bayesian classification for normal distributions, classification applications
Module 3: Algorithms and Nonlinear Classifiers (7 hours)
Linear discriminant-based algorithm: perceptron, perceptron algorithm
Support vector machines
Nonlinear classifiers, the XOR problem
Multilayer perceptrons
Backpropagation algorithm
Module 4: Unsupervised Learning and Ensemble Methods (8 hours)
Unsupervised learning
Clustering, examples, criterion functions for clustering
Proximity measures, algorithms for clustering
Ensemble methods: boosting, bagging
Basics of decision trees, random forest, examples
Module 5: Deep Learning Networks (7 hours)
Introduction to deep learning networks
Deep feedforward networks
Basics of convolutional neural networks (CNN)
CNN basic structure, Hyper-parameter tuning, Regularization - Dropouts
Initialization, CNN examplesassification, recognition and segmentation, speech recognition, automatic language translation and auto corrections, recommendation engines.
Bishop, C. M. (2006). Pattern Recognition and Machine Learning. New York: Springer.
Theodoridis, S., & Koutroumbas, K. (2003). Pattern Recognition. San Diego: Academic Press.
Hastie, T., Tibshirani, R., & Friedman, J. The Elements of Statistical Learning. Springer.
Duda, R. O., Hart, P. E., & Stork, D. G. Pattern Classification. New York: Wiley.
Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
Continuous Internal Evaluation: 40 marks
Quiz 1: 2.5 marks
Quiz 2: 2.5 marks
Seminar: 5 marks
Course Project: 20 marks
Internal Examination: 10 marks
End-semester Examination: 60 marks
As part of the course project, you are required to implement a machine learning algorithm using publicly available datasets. You are encouraged to select a topic from the provided list, although you are also free to explore any other area of interest. Please ensure you receive approval from the Course Instructor for your chosen topic. Additionally, you are required to submit a declaration confirming that your selected topic is distinct from the domain of your mini-project.Should the quality of your work meet the standards for academic publication, the Course Instructors may advise a joint submission for possible publication. This collaborative effort reflects the shared intellectual contribution and is a testament to the project's academic merit.
Suggested Project Topics:
Classification of phonological categories in imagined speech Reference Paper Dataset
Classification of motor imagery from EEG Reference Paper Dataset
Classification of imagined words from EEG Reference Paper Dataset
Modeling wine preferences by data mining from physicochemical properties Reference Paper Dataset
Breast cancer histopathological image classification using AlexNet Reference Paper Dataset
Music genre classification with convolutional neural networks Reference Paper Dataset
Sentiment classification system of twitter data for US airline service analysis Reference Paper Dataset
Classification of emotions from EEG Reference Paper Dataset
Online handwriting recognition system for Tamil Reference Paper Dataset
Real-time credit card fraud detection Reference Paper Dataset
Speech emotion recognition Reference Paper Dataset
Boston house price prediction using regression models Reference Paper Dataset
Breast cancer diagnosis Reference Paper Dataset
News Classification Reference Paper Dataset
Classification of sentiment polarity of cars and hotel reviews Reference Paper Dataset
Classification of fashion categories Reference Paper Dataset
Prediction of Titanic survival rate Reference Paper Dataset
Fundamentals of Linear Regression
Introduction to regression analysis
Linear regression model and its assumptions
Understanding error functions
Exploring Multivariate Regression
Extension from simple to multivariate regression
Application areas of multivariate regression
Addressing bias and variance
Introduction to Classification Techniques
Overview of classification in machine learning
Bayes’ decision theory basics
Discriminant functions and decision surfaces
Bayesian Classification Methods
Principles of Bayesian classification
Applying Bayesian classification to normal distributions
Real-world applications of Bayesian classification
The Perceptron Algorithm: A Linear Discriminant Approach
Concept and history of the perceptron
The perceptron learning algorithm
Limitations and use cases
Understanding Support Vector Machines (SVM)
Basics of SVM
Kernel trick and SVM optimization
SVM in classification tasks
Nonlinear Classifiers and the XOR Problem
Challenges with linear classifiers
Introduction to the XOR problem
Solutions via nonlinear classifiers
Dive into Multilayer Perceptrons (MLP)
Structure and functioning of MLP
Importance of backpropagation algorithm
MLP in complex problem-solving
Unsupervised Learning: An Overview
Distinction between supervised and unsupervised learning
Applications of unsupervised learning
Clustering Techniques and Their Applications
Understanding clustering and its criteria
Proximity measures and clustering algorithms
Examples of clustering applications
Ensemble Learning Methods: Boosting and Bagging
Concept of ensemble learning
Differences and similarities between boosting and bagging
Decision trees and random forests as examples
Introduction to Deep Learning Networks
Evolution and significance of deep learning
Key components of deep networks
Deep Feedforward Networks: Architecture and Applications
Understanding deep feedforward networks
Architectural nuances and application areas
Basics of Convolutional Neural Networks (CNN)
Introduction to CNNs and their unique architecture
Key operations in CNNs (convolution, pooling)
Hyper-parameter Tuning in CNNs
Importance of hyper-parameter tuning
Strategies for effective tuning
Regularization techniques, including dropouts
Initialization Techniques for Deep Learning
Role of initialization in model performance
Popular initialization methods and their impacts
CNN Applications: Image Classification and Beyond
CNNs in image classification
Extending CNN applications to recognition and segmentation
Speech Recognition with Deep Learning
Application of deep learning in speech recognition
Challenges and solutions in the field
Content (40 marks)
Accuracy (10 marks): Information is factually correct, well-researched.
Relevance (10 marks): Content is directly related to the seminar topic and objectives.
Depth (10 marks): Presentation covers the topic comprehensively, including background information and current trends.
Originality (10 marks): The presentation provides unique insights or a novel approach to the topic.
Organization (20 marks)
Structure (10 marks): Clear introduction, body, and conclusion; logical flow of ideas.
Pacing (10 marks): Time is well-managed, with neither rushed nor excessively slow segments.
Audio and Voice Delivery (20 marks)
Clarity (10 marks): Speaker articulates clearly, with good diction and appropriate volume.
Engagement (10 marks): Speaker uses tone variation and pauses effectively to maintain interest.
Visual Aids (10 marks)
Quality (5 marks): Slides or visual aids are legible, aesthetically pleasing, and free from excessive text.
Usefulness (5 marks): Visual aids enhance understanding of the topic and are relevant to the content discussed.
Understanding and Knowledge (10 marks)
Grasp of Topic (5 marks): Speaker demonstrates a strong understanding of the subject matter.
Responses to Hypothetical Questions (5 marks): Speaker anticipates and addresses potential questions in the presentation.
Technical Quality (10 marks)
Video/Audio Quality (10 marks): The audio is clear without background noise, and the video (if any visual elements are present) is steady and well-lit.
Please upload the slides and the videos here. Do not create a separate folder for each student. Format for filename: <RollNo>_<Name>_<Slides/Video>