| IS902E |
Linear Algebra |
9 |
Objectives: This is an introductory linear algebra course that aims to provide students with a solid foundation in mathematical concepts and techniques relevant to machine learning. |
- Vectors and Matrices: 2.5 hrs
- Inner Products and Norms: 2.5 hrs
- Linear Systems:2.5 hrs
- Sinirular Value Decomoosition:2.5 hrs
- Linear Independence, Basis, Dimension:2.5 hrs
- Solving Ax=b:2.5 hrs
- Eigenvalues, PD Matrices: 2.5 hrs
- Matrix Calculus:2.5 hrs
|
| EE901E |
Probability and Stochastic Processes |
9 |
Objectives:This course will focus on strengthening foundation of probability keeping its application in communications in mind. |
- Introduction to probability theory: Introduction to Probability and Probability Space, Properties of Probability Measure
- Random Variables: Distribution of Random Variables, CDF and PDF/PMF, Continuous and Discrete Random Variables, Examples of Random Variables
- Expectation: Moments, Variance, MGF
- Functions of Random Variables: Transformation of discrete and continuous random variables-
- Multiple RVs: Multiple RVs
- Sampling of random variable: empirical statistics using computer simulations
- Conditiona l Distribution:Conditional Expectation
- limit Theorems:Law of Large Number, Central Limit Theorem, Deviations
- Introduction to Random Processes:Distribution of Random Processes
- Random Proc:P.ssP.s via Linear Systems
|
| EE902E |
Digital Communications |
9 |
a. Objectives: |
- Introduction: An introduction to digital communication, Communication channels and models, Review of signals, Representation of lowpass and bandpass signals. Mathematical preliminaries: Signal space representation of waveforms, A brief introduction to random variables, Complex Random variables, A brief introduction to random processes.
- Memoryless Modulation: Memoryless modulation: An Introduction, Pulse Amplitude Modulation, Phase Shift Keying, and Quadrature Amplitude Modulation, Orthogona l, bi- orthogonal, and simplex signaling
- Modulation with memory: Modulation with memory: Continuous Phase Frequency Shift Keying, Continuous Phase Modulation
- Optimum receivers for AWGN channels: Optimum receivers for AWGN channels: Optimal Detection for a vector AWGN channel Waveform and vector AWGN channels, Optimal Detection for Binary Antipodal Signaling, Correlation Receiver, Matched Filter Receiver
- Probability of error computation: Probability of error computation for coherent detection: Optimal detection and error probability for ASK or PAM, and PSK signaling, Optimal detection and error probability for QAM signaling, Optimal Detection and error Probability for Orthogonal, Bi-Orthogonal and Simplex Signaling
- Noncoherent Detection: Noncoherent detection (Noncoherent detection of carrier modulated signals, Error Probability of Orthogonal signaling with noncoherent detection, Differential Phase Shift Keying) Detection of signals with memory Maximum likelihood sequence estimator, Viterbi Algorithm, Optimum receivers for CPM signal, 1.
- Communication over Bandlimited Channels: Design of Band-Limited Signals for No Intersymbol Interference-The Nyquist Criterion, Design of Band-Limited Signals with Controlled ISI-Partial-Response Signals, Data Detection for Controlled ISI, Optimum Receiver for Channels with ISI and AWGN
- Equalization: Linear Equalization, Decision-Feedback Equalization Predictive Decision-Feedback Equalizer, Equalization at the Transmitter: Tomlinson-Harashima Precoding,
- Adaptive Equalization: Adaptive Linear Equalizers. LMS, RLS
- Fading Channels: Characterization of Fading Multipath Channels, Diversity Techniques for Fading Multipath Channels, Digital Signalling over a Frequency-Selective, Slowly Fading Channel
- Multicarrier Modulation & MIMO: OFDM. PAPR reduction techniques, Issues due to frequency and timing offset, MIMO: Multiplexity, diversity, beamforming
|
| EE903E |
Information & Coding Theory |
9 |
Objectives: This course will help students answer two fundamental questions in
communications, namely, what is the ultimate data compression possible, and what is the maximum rate at which we can communicate reliably over a communication link.
Contents (preferably in the form of 5 to 10 broad titles): Introduction to Information & Coding Theory, Block to Variable length coding, Variable to block length coding, Variable to Variable length coding, Channel Capacity, Introduction to Linear Block codes, Bounds on the size of the code, Convolutional Codes, Decoding of convolutional codes, Turbo codes, LDPC codes, Polar codes |
- Entropy, Mutual Information: Introduction: Entropy, Relative Entropy, Mutual Information; Information Inequalities
- Block to Variable length coding: Block to variable length coding: Kraft’s inequality. Shannon-Fano coding, Huffman coding
- Variable to block length coding: Variable to block length coding: Tunstall coding; Block to block length coding, Typical sequences, Asymptotic Equipartition Property
- Variable to Variable length coding: Variable to variable length coding-I: Arithmetic codes, LZ77, LZ78, LZW algorithms, Channel Capacity
- Introduction to Linear Block codes:Types of error-correcting codes, FEC, ARQ, HARQ, Linear Block Codes, Examples of simple linear block codes. Construction, properties, and decoding of some popular block codes: Hamming codes
- Reed Muller Codes, Bounds on the size of the code: Reed-Muller codes, Bounds on size of the codes
- Convolutional Codes: Convolutional codes: Construction, structural properties, and distance properties
- Decoding of convolutional codes: Decoding of convolutional codes: Viterbi algorithm, BCJR algorithm
- Turbo codes: Turbo codes: Construction, distance properties, performance analysis, and iterative decoding.
- LDPC codes: Low-Density Parity Check Codes: Tanner graphs, LDPC code construction, belief propagation algorithm, and iterative decoding
- Polar codes: Polar codes: Encoding and decoding. Applications of modern coding techniques in 4G/5G communications
|
| EE904E |
Wireless Communications |
9 |
The course has both theoretical and practical flavors. It aims to explain the fundamental concepts and insights behind the development of modern 4G/ 5G wireless communication technologies. As part of the course, precise analytical models will be presented for various wireless systems followed by detailed performance analysis. The course intends to cover several key 4G/ 5G wireless technologies such as OFDM, MIMO, MIMO-OFDM in significant detail. This will also lay the foundation for advanced wireless communication techniques.
b. Contents (preferably in the form of 5 to 10 broad titles) |
- Wireless Communications and:AWGN channel modeling, SNR concept and BER performance for BPSK, QPSK and higher order modulations
- Digital communication system models: Fading channel models, BER analysis, Deep fade
- Diversity in Wireless Systems:Multiple antenna systems, Beamforming and diversity concepts
- Multiple Input Multiple Output (MIMO) Technology Receivers:MIMO Technology, Linear Receivers ZF, MMSE and performance
- MIMO Technology and Optimization, Space Time Codes:SVD, Precoding/ Combining in MIMO, Optimal Power Allocation, Space Time Block Codes
- Orthogonal Frequency Division Multiplexing (OFDM) technology:Single carrier vs. Multicarrier implementation, IFFT/ FFT receivers in OFDM, Cyclic prefix and circular convolution
- MIMO OFDM Technology:MIMO OFDM system model, transmission/ reception and receiver structure
- Wireless channel modeling:Wireless channel models, delay spread, frequency selective/ frequency flat channels, mobility and Doppler modeling
|
| EE928E |
Introduction to Machine Learning |
9 |
Objectives: This course aims to introduce the students to machine learning (ML) techniques used for various engineering applications. The lectures will focus on mathematical principles, and there will be coding-based assignments for implementation, introducing students to tools such as pytorch and sklearn. |
- Preliminaries, basics, applications: Linear Algebra, Probability
- Model Evaluation Measures: Precision, Recall, Accuracy, F-score
- Supervised Learning: Least square solutions, Logistic regression, SVM, kernel methods
- Unsupervised learning: Clustering, PCA, Distribution Learning
- Decision-based Ensemble Models: Decision tree, Random Forest, Bagging, Boosting
- Neural Networks: Perceptron, FFN, CNN, Gradient Descend Algorithm
- ML-ops, ML-at-scale: Federated Learning
|
| EE905E |
Computer & Wireless Networks |
9 |
Objectives: ● Provide a solid foundation in the principles of computer and wireless networks.
● Explain Internet architecture and the design of layered protocols across applications, transport, network, and link layers.
● Introduce performance analysis and security mechanisms in modern communication systems.
. |
- Introduction & Network Structure:Internet, services; Hierarchical organization; Packet vs. circuit switching; Delay, loss, throughput
- Network Architecture & App Basics:Internet 5-layer & OSI model; Encapsulation; Application architectures
- Application Layer Protocols: HTTP; FTP/SMTP/POP3/IMAP; DNS; Socket programming
- Transport Layer Fundamentals: UDP; Reliable data transfer; FSM models
- Reliable Transport Mechanisms: Stop-and-Wait; Go-Back-N; Selective Repeat; Intro to congestion control
- TCP Protocol: TCP header; 3-way handshake; Flow & congestion control (slow start, AIMD); Throughput; Wireshark
- Network Layer I: Forwarding & Addressing: IPv4, subnetting, CIDR; Forwarding tables; NAT; ICMP
- Network Layer II: Routing & IPv6: IPv6 basics; RIP, OSPF; BGP; Mobility
- Link Layer I: Wired Networks: Framing, error detection; ALOHA, CSMA; Ethernet; Switches
- Link Layer II: Wireless Networks: Wireless links; 802.11 MAC; CSMA/CA; Cellular mobility
- Network Security: Security goals; Attacks; Crypto tools; TLS, VPNs, IPsec
|
| EE906E |
Image Processing |
9 |
Objectives: This course aims to provide fundamental knowledge and practical techniques in digital image processing, covering theory, algorithms, and applications.
- Image formation and camera models
- Human visual system
- Image enhancement
- Color image processing
- Image restoration, denoising, quality assessment
- Image compression and morphological processing
|
- Introduction & Human Visual System: • Applications and scope of image processing • Structure and function of the human eye, visual perception basics • Brightness adaptation and contrast sensitivity
- Image Formation & Perception: • Image formation models and camera systems • Elements of visual perception and color perception • Radiometry, photometry, and imaging sensors
- Sampling and Quantization: • Digital image representation and resolution • Sampling theorem, aliasing, and anti-aliasing • Quantization techniques and visual effects
- Image Enhancement – Spatial Domain: • Gray-level transformations and histogram processing • Spatial filters forsmoothing andsharpening • Median, Gaussian, and Laplacian filters
- Image Enhancement – Frequency Domain: • Fourier transform and frequency representation of images • Low-pass, high-pass, and band- pass filtering • Homomorphic filtering for illumination correction
- Color Image Processing: • Color models: RGB, CMY, HSV, YCbCr • Color space conversions and enhancement • Applications in computer vision and multimedia
- Edge Detection & Feature Extraction: • Gradient-based edge detectors (Sobel, Prewitt, Canny) • Parametric vs. non- parametric edge models • Corner and blob detection methods
- Multi-Resolution Analysis & Segmentation: • Image pyramids and wavelet transforms • Thresholding and region-based segmentation • Watershed and edge- based segmentation
- Image Denoising & Quality Assessment: • Noise models and spatial/frequency filtering • ML-based denoising methods • Quality metrics: MSE, PSNR, SSIM
- Image Compression: • Lossless methods: Huffman, Run-Length, LZW • Lossy methods: JPEG, JPEG2000 •Transform coding and rate-distortion concepts
- Morphological Image Processing & Applications: • Binary morphology: erosion, dilation • Binary morphology: opening, closing
|
| EE907E |
Statistical Signal Processing |
9 |
The course aims to explain the fundamental principles and insights pertaining to statistical/ probabilistic signal analysis followed by their applications in various areas such as Signal Processing, Wireless Communication, Image Processing and also Machine Learning. As part of the course, detailed analytical models will be presented for various frameworks followed by detailed performance analysis. The course intends to cover several key concepts such as Maximum Likelihood, Minimum Mean Squared Error (MMSE), Kalman Filter, Adaptive Learning, Compressive Sensing, Sparse Learning and Probabilistic Graphical Models. Various applications will be covered such as Regression, Channel Estimation, MIMO, Auto-Regressive modeling, Recommender systems, Unsupervised Learning,
Probabilistic Clustering etc. |
- Maximum Likelihood Estimation: Likelihood cost function, Least Squares estimate, Error covariance
- Online estimation: Recursive Least Squares, Update Equations, Cost function and performance
- Bayesian Estimation: Principle of LMMSE, LMMSE estimator, Error covariance
- Adaptive Signal Processing: Introduction to adaptive signal processing, Steepest Descent, Algorithm and convergence analysis, Least Mean Squares (LMS) algorithm
- Iterative Estimation: Expectation- Maximization (EM) algorithm, Gaussian Mixture Models
- Compressive Sensing: Sparse estimation and sparse signal recovery, Orthogonal Matching Pursuit (OMP), Sparse Bayesian Learning
- MUSIC Algorithm: Introduction to array processing, Signal covariance matrix, Multiple Signal Classification
- Probabilistic Graphical Models: Introduction to Probabilistic Graphical Models, Bayesian networks, Bayesian inference and estimation over graphs
|
| EE908E |
Fundamentals of Data Science and Machine Intelligence |
9 |
Objectives: This course aims to equip students with the fundamental concepts and techniques central to the fields of exploratory data analysis, statistical inference, and machine learning, leading to machine intelligence. Students from all disciplines, both engineering and sciences, can develop proficiency in data analysis/visualization, statistical data analysis, machine learning algorithms, and machine learning tools, enabling them to obtain actionable insights from complex datasets in various domains by completing this course. Students will be exposed to the design and implementation of machine learning models and handling AI frameworks in Python via a course project. By the end of the course, students are expected to be able to design, implement, and evaluate machine intelligence in general, and thus prepare them for careers in data science, artificial intelligence, and related disciplines. The course is targeted at all engineering and science disciplines that wish to
understand the emerging and popular paradigm of Data Science and Machine Intelligence.
Contents (preferably in the form of 5 to 10 broad titles):
1. Foundational principles of data science and machine intelligence
2. Statistical data analysis, visualization, and inference
3. Regression analysis and modeling
4. Clustering, Decision Trees, PCA, ICA, Vector Quantization
5. Gaussian Mixture modeling
6. Artificial Neural Networks
7. Deep Neural Networks, Convolutional Neural Networks, Recurrent Neural Networks, and Autoencoders
8. Applications of machine intelligence in science and engineering
|
- Foundational principles of data science and machine intelligence: Introduction to Data Science (DS), and Machine Intelligence (MI) 2. Introduction to Algorithms, Models, Optimization Techniques for DS/MI. 3. Introduction to Supervised Learning, Unsupervised Learning, and other learningtechniques
- Statistical data analysis, visualization, and inference: 1. Statistical Data Analysis: Descriptive Statistics, Exploratory Data Analysis (EDA), Hypothesis Testing, Correlation and Covariance 2. Data Visualization: Histogram, Scatter Plot, Box Plot and other plots Model Visualization: Confusion Matrix, ROC curve Inference: Point Estimation, Confidence intervals, Bayesian Inference
- Regression analysis and modeling:1. Linear Regression Modeling 2. Non-Linear Regression Modeling 3. Logistic Regression 4. The Bias-Variance Decomposition 5. Bayesian Linear Regression
- Clustering, Decision Trees, PCA, ICA, Vector Quantization:1. Distance Measures 2. Clustering: K-Means, clustering variants, Hierarchical clustering 3. Decision Tree: Tree construction, Pruning 4. Random Forests: Ensemble learning, Out-of-bag error, Feature Importance 5. Bosting/Bagging
- Gaussian Mixture modeling:1. Probabilistic distance measures 2. Univariate and Multivariate distributions 3. EM algorithm 4. Gaussian Mixture modeling and applications in machine intelligence
- Artificial Neural Networks:1. Feed Forward neural networks 2. Network training 3. Error backpropagation, regularization.
- Deep Neural Networks, Convolutional Neural Networks, Recurrent Neural Networks, and Auto encoders: 1. Deep feed forward networks, Regularization and optimization for training deep models 2. Convolutional Neural Networks: Convolution operation, convolution and pooling, variants of convolution function. 3. Recurrent Neural Networks. Auto encoders and Applications
- Applications of machine intelligence in science and engineering: 1. Case studies of data science/machine intelligence in pure sciences 2. Case Studies of data science/machine intelligence in engineering
|
| EE909E |
Digital Switching |
9 |
Objectives: To discuss various kind of switches, their mathematical properties, design of circuit switches, packets switches, bounds on performance.
b. Contents (preferably in the form of 5 to 10 broad titles) |
- Lecture Organisation Week 1 Need for networks, Mesh topology; broadcast network, multichannel broadcast network; switched star network; complexity of number of links, total links lengths, number of transcievers; blocking and strictly non- blocking character of above network topologies; Backbone and access networks; Need of switched networks. Why switching?; optimization of cost; Access network, Backbone network; Process of making telephone call; manual telephony; mechanism of manual telephony; Generic exchange structure; Need for automatic exchanges; Strowger exchange - idea; uni-selector and two motion selector; crosspoint element;
- Week 2 Crossbar switch, Algorithm for crossbar switch operation, Various technologies for creating switching elements - digital logic gate based crossbar, optical coupler based switch, SOA switch, MEMS based switch; Sampling of voice in telephony; PCM encoding; frame rate- 125 μsecs frames; Insight into crosspoint complexity of two stage strictly non-blocking switch. Minimal number of crosspoint in two stage switch, Unique routing in two stage network, Three stage network, Alternate routing, Clos network definition
- Week 3 Cross point complexity reduction using multistage interconnection network; Clos network- three stage interconnection network; condition for strictly non-blocking switch; Crosspoint complexity of multistage strictly non-blocking network versus crossbar. Circuit switching and Packet Switching; M/M/1 queue - introduction; Poisson process, Exponential distribution, Markov chains, state probabilities, transition probabilities; Average queue length, average delay Little's theorm; Analysis of composite switch; states for simple queue (M/M/1/∞), states for composite switch, Probability of switch being in blocked state.
- Week 4 Call congestion and time congestion, The relationship in call congestion and time congestion for composite switch; Lee's approximation for Clos network; Karnaugh's approximation for Clos network; Call arrival rate in state , Computation of Probability of call loss ; for karnaugh approach; random route hunting, State space in the switch, definition of states in three stage network, Markov chain for composite switch; State probabilities for the composite switch; state probabilities; Jacobeous approaximation
- Week 5 Paull's matrix- conditions for legitimate Paull's matrix; Formal definition of strictly non-blocking network; Clos theorm- proof for condition of strictly non-blocking network; Slepian Duguid theorm- condition for rearrangeably non-blocking network; Proof of Slepian-Duguid theorm- using Paull's matrix; number of rearrangements needed; Paull's theorm- reducing the maximum rearrangements needed; looping algorithm for creating paths;
- Week 6 Widesense non-blocking switch example; proof of widesense non-blocking nature of the switch:
- Week 7 Recursive construction of strictly non- blocking Clos network; recursive construction of rearrangeable nonblocking Clos network; crosspoint complexity of recursively constructed rearrangeably nonblocking network. crosspoint complexity of recursively constructed strictly nonblocking network; discussion on various problems; Strictly non blocking network with reduced complexity- Cantor network;
- Week 8 Time, space, and code switching; RAM based implementation of time switch, Supermultiplexed switch; ST switch- control memory; TS switch; TST switch; control messages in the switches; Type-I, type-II and type-III messages, block diagram showing transfer of type-III messages from switch block control to switch control; control mechanism; Time multiplexed switch;
- Week 9 Packetization delay; packet transmission delay; jitter; routing, forwarding and switching; shortest path algorithm (Djikshtra's algo); Different switch implementation- mutistage, crossbar, shared memory, shared medium; virutal circuit switching; Input queueing versus output queueing- delay analysis; saturation throughput in input queueing systems, better throughput with dropping of backlogged packets.
- Week 10 Banyan networks; Delta-b network; Identity permutation; possible number of permutations for nonblocking connectivity patterns- probability of blocking state; non blocking routing for ordered inputs without gaps; Shufflenet, Proof of shufflenet's digit based routing property;
- Week 11 merge network, sorting network using merge network; nonblocking delta with sorting and identity permutation network combination. Bufferred Banyan network; Bufferred Banyan network and its analysis; Knockout switch architecture; Optical packet switching. VoIP – voice over IP; p2p media transport mechanisms; SIP signalling.
|
| EE910E |
Introduction to Reinforcement Learning |
9 |
Objectives: ● Introduce the core principles of reinforcement learning and the exploration– exploitation trade-off. ● Provide a foundation in sequential decision making through bandits and Markov decision processes. ● Familiarize students with key algorithmic approaches and modern extensions such as Deep RL.
|
- Foundations of RL: Agent–environment interaction; Rewards, returns, policies; Exploration vs. exploitation
- Bandit Models & Performance: Best Arm Identification; Regret minimization;
- Multi-Armed Bandit Algorithms: Action Elimination; UCB; Contextual bandits;
- Fundamentals of MDPs: MDP formulation; Value functions; Bellman expectation equations
- Dynamic Programming for MDPs: Bellman optimality; Policy evaluation & improvement; Policy/Value Iteration; Convergence
- Monte Carlo Methods: MC prediction & control; On- policy vs. off-policy; Importance sampling
- Temporal- Difference Learning: TD prediction; SARSA; TD(λ)
- Q-Learning & Variants: Q-learning; Double Q-learning; Comparing SARSA & Q-learning
- Function Approximation: MC & TD with approximation; Deep Q-Network (DQN)
- Policy Gradient Methods: Policy gradient theorem; REINFORCE algorithm
- Actor–Critic Methods & Applications: Actor–critic (PPO, A2C, DDPG); Applications
|
| EE911E |
Mathematical Structures of Signals and Systems |
9 |
Objectives: The course deals with mathematical methods for the discrete time detection of information transmitted through frequency selective channels. The proposed
methods are applicable to wireless technologies like 5G and beyond.
b. Contents (preferably in the form of 5 to 10 broad titles)
i. Review of digital signal processing (DSP).
ii. Detection of information transmitted through frequency selective channels:
I. Receivers based on equalization.
II. Maximum likelihood (ML) receivers.
III. Orthogonal frequency division multiplexing (OFDM).
IV. Multicarrier communications.
|
- Review of digital signal processing (DSP): Sampling theorem, discrete-time Fourier transform (DTFT), z- transform, discrete Fourier transform (DFT), fast Fourier transform (FFT),decimation and interpolation.
- Receivers based on equalization: Symbol-spaced – infinite length, finite length, least mean squared algorithm, steepest descent algorithm, fractionally- spaced equalizers,linear prediction, decision feedback equalizers.
- Maximum likelihood (ML) receivers: symbol-spaced, fractionally spaced ML receivers, symbol error rate analysis.
- Orthogonal frequency division multiplexing(OFDM).
- Multicarrier communications: OFDM offset quadrature amplitude modulation (OQAM), filter bank multicarrier (FBMC), universal filtered multicarrier (UFMC).
|
| EE912E |
Detection and Estimation Theory |
9 |
Objectives: To make the students familiar with the basic concepts of hypothesis testing and parameter estimation.
b. Contents (preferably in the form of 5 to 10 broad titles) |
- Binary and M-ary Hypothesis Testing: Bayesian hypothesis testing, sufficient statistics, ROC, NP test, Minimax testing
- Tests With Repeated Observations: Sequential testing, optimality of SPRT.
- Parameter Estimation Theory: Bayesian estimation, MSE estimation, ML estimation, CRLB
- Composite Hypothesis Testing: UMP test, GLRT
- Detection of Signal in Gaussian Noise: Energy test, Matched Filter test
|
| EE913E |
Computer Vision and Deep Learning |
9 |
Course Description: This course provides a comprehensive introduction to computer vision with deep learning, spanning both foundational methods and recent advances. Topics include image formation, classical feature extraction, and modern architectures such as convolutional neural networks and vision transformers. Core tasks such as classification, detection, and segmentation are complemented by modules on representation learning, generative models, 3D vision, video understanding, and multimodal approaches. The course also addresses applications in domains such as medical imaging, autonomous driving, and robotics, as well as challenges related to efficiency, robustness, and ethical considerations.
Students will gain the knowledge and skills to design, implement, and evaluate real-world vision systems.
a. Objectives: ● Understand the foundations of computer vision and deep learning, including classical methods, modern architectures, and core visual recognition tasks.
● Apply advanced models such as CNNs, transformers, and representation learning techniques to solve problems in image, video, and 3D vision.
● Analyze and develop generative and multimodal approaches (GANs, diffusion models, vision–language systems) for content creation and cross-modal understanding.
● Evaluate, optimize, and critically reflect on computer vision systems with respect to efficiency, robustness, ethical considerations, and real-world applications.
b. Contents (preferably in the form of 5 to 10 broad titles) |
- Foundations of Computer Vision: Introduction to CV; Image formation, color spaces, filtering, features; Transition from classical CV to deep learning
- Deep Learning for Vision: Neural networks, backpropagation, optimization; CNN architectures (AlexNet → ResNet, modern CNNs); Transfer learning; Regularization, augmentation, training tricks; Vision transformers
- Core Vision Tasks: Image classification; Object detection (R- CNN, Faster R-CNN, YOLO, transformers); Semantic segmentation (FCNs, U-Net); Instance and panoptic segmentation; Evaluation metrics and benchmarks
- Representation Learning:Self-supervised and contrastive learning; Vision transformers (ViT, Swin); Multimodal representation learning (CLIP and related models)
- Generative Computer Vision: Generative models overview (VAEs, GANs); GAN architectures and training challenges; Diffusion models (DDPM, Stable Diffusion); Text-to-image generation (DALL·E, Imagen); Video and 3D generation
- 3D Vision & Video Understanding: Multi-view geometry and depth estimation; 3D reconstruction and NeRFs; Video understanding (action recognition, tracking); Video generation and prediction
- Multimodal & Specialized Applications:Image captioning and VQA; Vision– language foundation models (BLIP, Flamingo, LLaVA); Applications in medical imaging, autonomous driving; AR/VR and robotics
- Frontiers & Ethics in Vision AI: Model efficiency (pruning, quantization, distillation); Robustness, adversarial attacks, explainability; Ethics, bias, and the future of vision AI
|
| EE914E |
Convex Optimization |
9 |
Objectives: This course provides a comprehensive foundation in convex optimization, covering theory, algorithms, and applications in engineering.
It develops the ability to formulate, analyze, and solve optimization problems using convex sets, functions and mathematical techniques.
- Mathematical foundations - Convex sets and functions - Optimization principals and methods - Duality and conjugate analysis - Applications in engineering |
- Introduction & Mathematical Foundations: ● Applications and scope of convex optimization ● Matrix operations and manipulations ● Eigenvalues and eigenvectors in optimization
- Convex sets: ● Definition and properties of convex sets ● Hyperplanes, half-spaces, and supporting hyperplanes ● Intersection and separation theorems
- Convex functions: ● Definition and properties of convex functions ● Types: strictly convex, strongly convex, concave ● Operations that preserve convexity
- Optimization:● Formulating optimization problems ● Feasibility and optimality conditions ● Gradient and sub-gradient methods
- Linear optimization: ● Linear programming formulation ● Simplex method overview ● Duality in linear programming
- Quadratic optimization: ● Unconstrained and constrained quadratic programming ● Karush–Kuhn–Tucker (KKT) conditions ● Applications in control and machine learning
- Duality & Conjugate Functions:● Primal and dual problems ● Weak and strong duality ● Fenchel conjugates and applications
- Semidefinite Programming: ● Definition and matrix inequalities ● Standard and dual forms of SDP ● Applications in control, signal processing, and ML
- Geometric Programming: ● Posynomials and monomials ● Logarithmic transformation for convexity ● Applications in engineering design
- Special Topics: ● Compressed sensing and sparse recovery ● Support Vector Machines (SVM) optimization
- Applications: ● Review of convex optimization techniques ● Case studies in communications, ML, and signal processing
|
| EE915E |
MIMO Wireless Communications |
9 |
Course Description: This course will cover state-of-the-art multiple-input multiple-output (MIMO) wireless transmitter and receiver designs which are being used in the current 4G/5G and in the emerging 6G cellular systems.
Objectives: The students will learn about the a) challenges involved in developing and incorporating MIMO designs in practical wireless systems; and b) performance improvements MIMO provide over single-antenna designs in different operating scenarios.
- Mathematical foundations - Convex sets and functions - Optimization principals and methods - Duality and conjugate analysis - Applications in engineering |
- Basics of Information Theory:Mutual information, capacity of discrete and continuous channels
- Transceiver design for fixed MIMO channel: Maximal ratio combining, maximal ratio transmission, SVD-based transceiver
- Modelling of LoS MIMO channels: Array response vector, DFT based codebook, capacity discussion
- Transceiver design for fading MIMO channel:Outage and Ergodic capacity for slow and fast fading channels, open loop transmission, ZF, MMSE, receiver design
- Transceiver design for Multi-User MIMO: Uplink and downlink transceiver design, duality,
- Transceiver design for multi-cell multi- user MIMO: Multi-cell Massive MIMO, Multi-cell interference and
|
| EE916E |
Peer-to-peer networks |
9 |
Objectives: Objectives: Basics of Peer-to-peer networks, implementation of various services on p2p networks. |
- Proposed content for discussion.
- Week 1 - P2P system design issues. - Structured and unstructured query networks.
- Week 2 - Various DHT algorithms, their Characterizations - Security setup
- Week 3 - Transport network design basically firewalls punching, proxies.
- Week 4 - Multilayer DHT for different services in same system.
- Week 5 - Distributed server less search engine
- Week 6 - Distributed P2P resilient storage - Distributed File System/Universal File System
- Week 7 - Torrent based system - Distributed mailing/ messaging/ mailing list system
- Week 8 - Distributed compute
- Week 9 - P2P live streaming system. - Brihaspati4 architecture. - P2Psip – RFC from www.ietf.org
- Week 10 - The Onion Routing (TOR) for anonymity
- Week 11 - Darknet architecture (anonymous services) - Blockchain and other alternative architectures - Delay Tolerant Networking – networking without backbone
|
| EE917E |
5G Wireless Technologies |
9 |
Course Description: The course will present an in-depth analysis of several key 5G wireless technologies such as Massive MIMO, mmWave MIMO, Filter Bank Multi-Carrier (FBMC), Non-Orthogonal Multiple Access (NOMA), Full-Duplex (FD) etc. It is also intended to cover other advanced wireless technologies such as Cooperative Communication, Cognitive Radio, Multi-User MIMO and more. Finally, students working in groups of two are expected to prepare a term paper that will focus on an in-depth study and analysis of any cutting edge 5G wireless technology of their choice.
|
- Introduction to 5G Wireless Networks and Technologies:Evolution of Wireless Cellular Technologies 3G/ 4G/5G
- Introduction to MIMO:Performance of Multi- Antenna and MIMO, Performance of Multi- Antenna and MIMO
- Massive MIMO:Signal Processing for massive MIMO, Channel Estimation for massive MIMO, Spatial Modulation of Massive MIMO
- mmWave Wireless Systems, Hybrid Processing for mmWave MIMO:Introduction to mmWave MIMO, Properties and Modeling of mmWave Wireless Channels, Channel Estimation in mmWave Wireless Systems, Optimal RF/ Baseband Precoders and Combiners for mmWave MIMO
- Non-Orthogonal Multiple-Access (NOMA):Introduction to NOMA Wireless Systems, System Model and Decoding for NOMA Systems
- Performance Analysis of NOMA:Optimal Performance of NOMA, Uplink and Downlink NOMA Systems
- Cooperative Wireless Communication:Protocols for Cooperative Wireless Communication, Decode-and-Forward (DF), Performance Analysis of DF
- Cognitive Radio Systems:Introduction to Cognitive Radio, Spectrum-Sensing, Optimal Power Allocation
|
| EE918E |
Analysis of Modern Wireless Networks |
9 |
Objectives: This course will cover tools from stochastic geometry to model and analyze modern wireless systems being used in 4G, 5G and Beyond-5G systems. After completion of the course, the students should be able to apply mathematical tools from stochastic geometry in their own research to analyze modern wireless systems. |
- Introduction to modern communication networks:Introduction and evolution of communication systems; Modeling and analysis issues in modern communication systems.
- Introduction to stochastic geometry:Why use stochastic geometry?; Its applications and validation Point processes (PPs) and their characterization.
- Poisson point process:Poisson Point Process (PPP); Types and properties of PPP; Thinning, displacement and superposition of PPPs ; Laplace functional and PGFL of PPP; Campbell theorem Marked point process
- Performance of an ad-hoc network:Interference characteristics; 2) Transmission capacity 3) SINR distribution.
- Downlink/Uplink cellular system:1) Interference characteristics; 2) SINR and rate coverage; 3) Impact of fading, shadowing; and 4) MIMO.
- Palm Calculus:Reduced palm distribution; 3) Marked point processes; 4) Campbell Mecke theorem, Slivnyak theorem
- HetNets:Introduction to modern HetNets; Performance with various cell association rules; HetNet MIMO
- Boolean Models
- 6G NetworksMmWave and THz networks,
|
| EE919E |
Speech Signal Processing |
9 |
Objectives: This course will deal with both theory and practical aspects of Speech signal processing. The course requires basics understanding of digital signal processing and probability theory. |
- Introduction to modern communication networks:a) The physiological model of speech production b) The mathematical (source-system) model of speech production c) Relation between (a) and (b) d) Physiological and mathematical basis of categorization of speech sounds
- Introduction to stochastic geometry:Why use stochastic geometry?; Its applications and validation Point processes (PPs) and their characterization.
- Basic Signal processing techniques for speech recognition:a) Discrete time speech signals, relevant properties of the fast Fourier transform and Z- transform for speech processing. Convolution, filter banks, and analytical pole-zero modeling of the speech signal b) Spectral estimation of speech using the Discrete Fourier transform c) Pole-zero modeling, linear prediction (LP) analysis, perceptual linear prediction (PLP), analysis of speech d) Homomorphic speech signal deconvolution, real and complex cepstrum, application of cepstral analysis to speech signals
- The speech recognition front end and pattern comparison techniques:a) Mel frequency cepstral co-efficients (MFCC), MVDR-MFCC, RASTA-PLP cepstral co-efficients. b) Issues in feature vector extraction for speech recongition, Static and dynamic feature vectors for speech recognition, robustness issues, discrimination in the feature space, feature selection c) Log spectral distance, cepstral distances, weighted cepstral distances, distances for linear and warped scales
- Statistical models for speech recognition:a) Vector quantization models for speech and speaker recognition b) Gaussian mixture modeling for speaker, language and speech recognition c) Hidden Markov modeling for isolated word and continuous speech recognition
- Speech Recognition in practice: Using HTK for speech recognition
|
| EE920E |
Quantum Computing and Communication |
9 |
Objectives: This course will discuss quantum computing and communication. |
- Introduction:Introduction, Review of Linear algebra, Classical vs Quantum Behavior, Applications
- Quantum Postulates:Quantum postulates: Hilbert Space, time evolution, measurement, wavefunctions, Bloch sphere, POVM, Time Evolution of Pure State, Born Rule, State discrimination, Tensor Product, and Composite systems, Quantum Entanglement, Measurements (Partial, Destructive), bell states, quantum CNOT gate, Dense Coding, Teleportation, No-Cloning,
- Density Matrices and Mixed states:Density matrix, Ensemble, Time Evolution, Purification,
- Quantum Computing:Classical vs Quantum Computing, Quantum Gates and Universality, Quantum Algorithms: Groover’s, Simon’s, Shor’s, Deutsch–Jozsa Algorithm Conditional gates, EPR Paradox, Bell’s Theorem, CHSH Inequality, Generalized Measurement, Kraus Operator,
- Quantum Information theory:Classical Information Review, Quantum Information: entropy, Joint entropy, relative entropy, Typical sequences, Shannon noiseless coding theorem and quantum version, Distance metrics: trace distance and fidelity Holevo bound, entanglement fidelity,
- Quantum Communications: Quantum noise, quantum Channels, examples. communication over noisy quantum channels Quantum communications, Quantum Key Distribution (QKD), Continuous Variable QKD
|
| EE921E |
Coherent Imaging |
9 |
Objectives: Gaining basic understanding and insights about coherent imaging methodology in optics. Exposure to state of the art coherent imaging methods and their
diverse applications in optical metrology. |
- Introduction to Coherent Imaging:Optical Imaging, Wave optics and Maxwell’s Equations, Principles of Image formation, Coherent Imaging modality and significance, Applications: Nanoscale surface metrology, 3D shape detection, Non-destructive precision testing, Automated defect detection, Label-free biomedical imaging
- Mathematical Foundations:Introduction to Fourier transform Fourier transform properties Convolution and Correlation Hilbert transform and Analytic signal Linear System Theory Frequency response
- Light Wave Propagation:Electromagnetic Wave equation Helmholtz model Diffraction Theory Fresnel and Fraunhofer Approaches
- Fourier Picture of Propagation:Plane wave decomposition of signal Frequency domain analysis Angular Spectrum Transfer function for free space propagation
- Imaging Lens Analysis:Thin lens design Lens transmission model 2f lens Fourier System 4f lens Imaging System
- Optical Image Filtering: Fundamentals of spatial filtering Low pass filter High pass filter Convolution and correlation filters
- Optical Coherence:Random nature of light wave Spatio-temporal field correlations Wiener Khinchine Theorem and power spectral density Temporal and Spatial Coherence Coherent versus incoherent imaging
- Imaging System Characterization:Abbe’s imaging model Point spread function Imaging resolution Image deconvolution and inverse filter
- Optical Speckles:Coherent light and speckle phenomenon Mathematics of speckle formation Speckle characterization Speckle contrast and size
- Coherent Imaging Techniques and Applications: Phase contrast microscopy Digital speckle imaging Digital holography structured illumination imaging Applications in precision metrology,non-destructive testing and evaluation, and microscopy
|
| EE922E |
Natural Language Processing |
9 |
Objectives: To study Natural Language Processing (NLP) and learn statistical and deep learning algorithms for modeling and automated understanding of language. |
- Week 1: Introduction to Natural Language Processing:1. Why is NLP Hard? 2. Zipf’s Law 3. Levels of Language Processing
- Week 2: NLP Pipeline:1. Processing text using NLP Pipeline 2. NLP Tasks and Applications 3. Sub-tokenization algorithms
- Week 3: Text Prediction:1. Introduction and Examples 2. Prediction Framework 3. Feature Function and Feature Engineering 4. Prediction Model
- Week 4: Loss Function in Prediction Framework:1. Concept of loss function and regularization 2. Regression Loss Functions 3. Energy Based and Margin Based Loss Functions
- Week 5: Loss Function in Prediction Framework Continued: 1. Concept of Data Likelihood and Maximum Likelihood estimate 2. Classification Loss Functions 3. Contrastive Loss Functions 4. Ranking-based Loss Functions 5. Softmax Function
- Week 6: Classical Prediction Models: 1. NLP Model Loss Optimization 2. Stochastic Gradient Descent (SGD) and Other optimizers 3. Naïve Bayes Model 4. Expectation Maximization algorithm for Clustering
- Week 7: Introduction to Deep Learning:1. Introduction to Neural Networks (NN) 2. Computational Graphs 3. Types of Neural Networks and inductive biases 4. NN Practicalities 5. Convolutional Neural Networks (CNN) for Text Prediction
- Week 8: Sequence Modeling:1. Neural Sequence Models: RNNs, LSTM, GRU 2. Attention Mechanism
- Week 9: Word Representation:1. Distributed Representations 2. Word Embeddings: Types, training, models, contextualization
- Week 10: Transformers: 1. Self-Attention 2. Architecture and Training 3. Pre-trained Language Models: BERT, GPT, etc
- Week 11: Large Language Models (LLMs): 1. Language Modeling (LM) Task 2. Comparison of classical LM and LLMs 3. Emergent Capabilities in LLM 4. Key Techniques behind LLMs: Scaling, Training, Alignment Tuning, etc. 5. How to Build LLMs? 6. Using LLMs: Parameter Efficient Fine-tuning (PEFT)
|
| EE923E |
AI in Healthcare |
9 |
Course Description: This course provides a comprehensive overview of the use of artificial intelligence (AI) in healthcare, combining foundational concepts with applications in medical imaging, clinical data analysis, and personalized medicine. Students will learn machine learning and deep learning techniques tailored to healthcare challenges, explore state-of-the-art applications such as medical image analysis, electronic health record (EHR) mining, drug discovery, and predictive modeling, and critically assess ethical, regulatory, and fairness considerations in deploying AI systems for healthcare. By the end of the course, students will gain both theoretical knowledge
and practical skills to design, evaluate, and responsibly apply AI methods in healthcare contexts.
Objectives: By the end of this course, students will be able to:
● Understand core AI and machine learning methods relevant to healthcare data.
● Apply deep learning techniques to medical imaging, clinical data, and predictive healthcare tasks.
● Evaluate AI solutions in terms of accuracy, interpretability, robustness, and fairness in healthcare contexts.
● Critically analyze ethical, regulatory, and societal implications of AI in healthcare.
|
- Introduction & Foundations:Overview of AI in healthcare; Types of healthcare data (EHR, imaging, genomics); Basics of machine learning for healthcare
- Machine Learning in Healthcare:Supervised/unsupervised methods for healthcare data; Feature engineering in clinical data; Predictive analytics for patient outcomes; Time-series analysis of medical records; Evaluation metrics for healthcare tasks
- Deep Learning in Medical Imaging:CNNs for imaging; Segmentation and detection in radiology/pathology; Transfer learning in medical imaging; 3D imaging (CT, MRI); Multimodal imaging; Case studies (X-ray, MRI, histopathology)
- Natural Language Processing in Healthcare (4 lectures):Clinical text mining; Named entity recognition (diseases, drugs, symptoms); Clinical summarization; Large language models in healthcare
- Genomics and Precision Medicine: AI for genomics data; Biomarker discovery; Personalized treatment recommendation
- Generative AI in Healthcare: Generative models for medical imaging (GANs, diffusion); Synthetic data generation for privacy; Drug discovery with AI; Protein folding (AlphaFold)
- Healthcare Applications & Systems:AI in diagnosis, prognosis, and triage; AI in robotics and surgery; AI in telemedicine and wearable devices; Decision support systems; Real-world deployment case studies
- Challenges, Ethics & Regulation:Explainability and interpretability; Bias and fairness; Data privacy and security; Regulatory and ethical considerations
|
| EE924E |
Unsupervised Learning |
9 |
Objectives: This course introduces key concepts and techniques in unsupervised machine learning, including clustering, dimensionality reduction, and probabilistic modeling. Students will explore methods such as K-means, spectral clustering, graphical models, and mixture models with EM. The course also covers advanced topics like approximate inference and deep generative models including VAEs and GANs. |
- Clustering:Introduction and K-means Clustering Hierarchical and Spectral Clustering
- Dimension Reduction:Linear, Nonlinear Methods and 2D Visualization
- Probabilistic ML:Intro to probabilistic distribution and density estimation
- Graphical Models:Probabilistic Graphical Models, Bayesian Networks, Markov Random Fields
- Mixture Models: Gaussian Mixture Model, Expectation Maximization, Alt-Opt method
- Approximate Inference: Variational Inference, Markov Chain Monte Carlo
- Deep Generative Models:Variational Autoencoders, Generative Adversarial Networks
|
| EE925E |
Modern Technologies for 5G and Beyond |
9 |
Objectives: |
- Evolution of cellular networks, Exploring higher bands: mmWave:Introduction, mmWave: Propagation channel, physical layer issues, MAC layer, and network layer issues
- THz networks:Introduction, Channel Modeling, Spectrum management, beamforming, Applications: THz Sensing
- NTN, UAVs, and satellite communications:Introduction, NTN in future networks, multi-layer NTN architecture, UAV: Deployment, user association
- OTFS:Introduction, Delay-Doppler Modulation, Transceiver signal processing, waveform
- NOMA: Introduction, power-domain NOMA, code-domain NOMA, cooperative NOMA, applications
- Multicell coordination, COMP, cell-free systems:Introduction, CoMP in LTE-A, Cloud- based Cell-free distributed MIMO systems, Deployment of cell-free systems
- Reconfigurable intelligent surfaces:Introduction, Multi-function IRS, enabling technologies, fundamental issues, applications
- Cognitive networks, spectrum sharing:Introduction, architecture, spectrum sensing, resource allocation
- Integrated sensing and communication: Introduction, Channel Modeling, Multi- Beam ISAC, waveform Design,
- Nano-networks, molecular communication:Introduction, Internet of Bio-nano things, signal detection, timing alignment, MIMO in molecular communications,
- Visible light communication, Semantic communication:Introduction to VLC, Channel Models, Hybrid VLC and RF systems, VLC- based networking, Introduction to Semantic communication, Architecture, Cognitive semantic communications
|
| EE929E |
AIML Projects with real-world datasets |
9 |
Course Description: Students will participate and successfully complete several PYTHON-based projects and case-studies on key AI/ ML techniques. Students will also develop the skills to effectively use integrated development environments (IDEs) in PYTHON for tackling more extensive AI/ ML projects |
- Introduction to PYTHON:ML Packages, Data compression, Principal Component Analysis
- Linear Regression:Regression concepts and mathematical model. Least squares IRIS Dataset Regression Boston Housing Price Analysis
- Logistic Regression:Logistic regression modeling Purchase/ Shopping Data project application for Wine quality dataset
- Support Vector Machines:Hyperplanes, distance, SVM Optimization modeling Breast Cancer Dataset Analysis IRIS Data Set
- Naïve Bayes: Probabilities, Likelihood, Posterior Probability, Examples Purchase Dataset Project Wine quality data set project
- Linear Discriminant Analysis:Multivariate Gaussian Modeling, LDA Model Discriminant Based Data Classification using IRIS Data Set
- Decision Tree Classifiers (DTC):Information theory, Information gain concept, Restaurant waiting example Purchase Logistic Data Set project
|
| EE926E |
Machine Learning for Wireless Communications |
9 |
Course Description: Machine learning is now routinely being applied to design state-of-the- art 5G and 6G wireless systems. The objective of this course is to bridge the gap between theory of machine learning and its applications to 5G systems
Underlying machine learning concepts will be taught in the class. Students will then design and implement various applications of machine learning for 5G wireless systems in MATLAB. The course will therefore involve MATLAB coding component, which will also be considered for evaluation. |
- Linear Modeling: A Least Squares Approach:Linear modeling Generalization and over fitting Regularized least squares Wireless application - MIMO zero-forcing receiver design
- Linear Modelling: A Maximum Likelihood Approach:Errors as noise – thinking generatively Maximum likelihood Bias-variance trade-off Effect of noise on parameter estimates Wireless application - MIMO MMSE receiver design
- The Bayesian Approach to Machine Learning:Exact posterior Marginal likelihood Hyperparameters Bayesian Inference Non-conjugate models Point estimate – MAP solution Laplace approximation Wireless application - Massive MIMO channel estimation
- Classification:Probabilistic classifiers – Bayes classifier, logistic regression Non Probabilistic classifiers- K-nearest neighbors Discriminative and generative classifiers Wireless application - Detection in digital communication systems
- Sparse kernel machines: Support vector machines (SVM) Sparse Bayesian learning (SBL) Wireless application SVM for beamforming and data detection in millimeter wave systems SBL for channel estimation in massive MIMO
- Clustering:General Problem K-means clustering Gaussian mixture models (GMM) EM algorithm – MAP estimates, Bayesian mixture models Wireless application - Clustering for massive MIMO system using K means and GMM
|
| EE927E |
Simulation Techniques in Modern Wireless |
9 |
Objectives: This course will enable students to evaluate real-world modern communication systems via numerical methods. Numerical evaluation is a quick way to evaluate complex systems where analysis is difficult. Most of the telecommunication industry relies on simulations to test their methodology. Academicians use simulation to validate their analysis and extend their results for complex systems. This course will focus on simulation methodologies in the field of communication with a great focus on their actual implementations. The course is balanced version of theory and implementation. It would discuss fundamental tools in numerical techniques and its application to communications. |
- Introduction to course:Introduction to Simulation Methodology
- Signal/Systems:Representation of Signals, Representation of Systems
- Random Signals:Random Variables, Random Signals
- System Dynamics:Numerical Techniques, Differential Equations and Markov Chains
- Monte-Carlo Simulations: Monte-Carlo Simulations
- Link Level Simulation:Simulation of a communication channel, Wireless Channel
- Advanced Link Level Simulation: Advanced Modulation, MIMO
- System Level Simulations:Mobile Ad-hoc Networks, Cellular Networks, Millimeter wave and THz Networks
|