Special Topics

Special Topics courses cover some of our most innovative and promising research directions.  They are often prototypes of new courses that we are developing.

Special Topics courses offer variable course content, so each semester's offerings are unique.

Winter 2026

We aren't offering any courses in Winter 2026.

Spring 2026

CSE 510 Machine Learning: Basics, Optimization and Theory (Lecture)
Section: JI
Instructor: Kaiyi Ji
Description: This course revisits the foundations of machine learning in the era of large models and massive data. We will explore perspectives that challenge conventional conclusions—such as the relationship between generalization, sharpness, and flatness, as well as the roles of overparameterization and optimization dynamics. The course will also cover recent advances in large-scale optimization methods, including Adam, Shampoo, and Muon, along with emerging learning paradigms such as meta-learning, continual learning, reinforcement learning, parameter-efficient fine-tuning of large language models (e.g., LoRA), and direct preference optimization (DPO). A solid theoretical background in mathematics, statistics, and machine learning is required; students are encouraged to complete the relevant prerequisite courses beforehand.
Instruction Mode: In person
Class #: 23161
Dates: 01/21/2026 - 05/05/2026
Days, Time: W, 2:00PM-4:50PM
Location: Bell 325, North Campus
Credit Hours: 1-3
Enrollment: 2/ 9 (Active)
Info:
CSE 610 Machine Learning: Basics, Optimization and Theory (Lecture)
Section: JI
Instructor: Kaiyi Ji
Description: This course revisits the foundations of machine learning in the era of large models and massive data. We will explore perspectives that challenge conventional conclusions—such as the relationship between generalization, sharpness, and flatness, as well as the roles of overparameterization and optimization dynamics. The course will also cover recent advances in large-scale optimization methods, including Adam, Shampoo, and Muon, along with emerging learning paradigms such as meta-learning, continual learning, reinforcement learning, parameter-efficient fine-tuning of large language models (e.g., LoRA), and direct preference optimization (DPO). A solid theoretical background in mathematics, statistics, and machine learning is required; students are encouraged to complete the relevant prerequisite courses beforehand.
Prereqs: Introduction to machine learning
Instruction Mode: In person
Class #: 23019
Dates: 01/21/2026 - 05/05/2026
Days, Time: W, 2:00PM-4:50PM
Location: Bell 325, North Campus
Credit Hours: 3
Enrollment: 2/10 (Active)
Info:
CSE 610 Systems Support for Modern Applications (Lecture)
Section: LU
Instructor: Haonan Lu
Description: In this course, we will discuss the distributed system solutions for modern AI/ML applications. We will cover system challenges in training frameworks and inference architectures. This course is a mix of lectures, paper reading, and presentations.
Prereqs: CSE586
Instruction Mode: In person
Class #: 23023
Dates: 01/21/2026 - 05/05/2026
Days, Time: R, 11:00AM-1:50PM
Location: Baldy 109, North Campus
Credit Hours: 3
Enrollment: 5/10 (Active)
Info:
CSE 610 Distributed Quantum Computing (DQC) (Lecture)
Section: QIAO
Instructor: Chunming Qiao
Description: Quantum Computing is expected to be the next wave beyond Al/ML that will lead to unprecedented advancement in science and engineering. There are several main challenges to overcome before quantum computing can be practically useful. These include the small number of qubits (up to a few thousand) available in a quantum processing unit (QPU) and their low fidelity. A promising way to address the challenges is to network several small QPUs together using a quantum data network (QDN) and employ quantum error correction (QEC). This course will first introduce basic quantum computing, communication and networking concepts, and then delve into research issues in distributed quantum computing (DQC), including but not limited to those related to QDNs and QEC. Given that DQC is an emerging topic, the students are expected to possess a strong interest in research, and the ability to read and present research papers. Although no prior knowledge of quantum physics or quantum mechanics is needed, the students should have a background in linear algebra, algorithms, and networks. The students may be asked to take CSE439/539 (Quantum Computation), CSE489/589 (Modern Networking Concepts), or similar courses concurrently. Knowledge in distributed systems, datacenters, optimization, and algorithms would be a plus.
Prereqs: Although no prior knowledge of quantum physics or quantum mechanics is needed, the students should have a background in linear algebra, algorithms, and networks. The students may be asked to take CSE439/539 (Quantum Computation), CSE489/589 (Modern Networking Concepts), or similar courses concurrently. Knowledge in distributed systems, datacenters, optimization, and algorithms would be a plus.
Instruction Mode: In person
Class #: 22474
Dates: 01/21/2026 - 05/05/2026
Days, Time: R, 1:00PM-3:50PM
Location: Davis 113A, North Campus
Credit Hours: 3
Enrollment: 2/10 (Active)
Info:
CSE 610 AI Hardware Infrastructure: Systems and Architecture (Lecture)
Section: SRID
Instructor: Ramalingam Sridhar
Description: This course focuses on AI Infrastructure from the Systems and Architecture perspective. AI Hardware architecture plays a huge role in supporting the exploding AI Applications in all diverse domains. Many large companies such as nVidia, AMD, Qualcomm, Google, Amazon, Microsoft, IBM, TSMC and Intel, many startups, in hardware accelerators, memory systems programmable logic and neuromorphic computing. all play a strong role in enabling AI Applications in cost and power effective real-time environments. The course is targeted to advanced graduate students with some knowledge in computer architecture or related areas. Topics covered will include Domain specific architecture, Neuromorphic computing, Edge computing, memory architectures, reconfigurable computing, hardware accelerators for domain specific tasks, and machine learning. Design approaches will include high performance computing, power aware solutions and reconfigurable computing. Special approaches are needed for developing memory architectures to deal with very large amount of data, through high performance memory, in-memory computing, novel memory bus structures and memory technologies. Reconfigurable architectures play a significant role in easy manipulation of the systems in diverse environment needs. The course will include a comprehensive group project in which each student can contribute from his/her domain of knowledge, developing the solution using a holistic approach. Please email rsridhar@buffalo.edu for more details and requirements.
Notes:
Instruction Mode: In person
Class #: 23947
Dates: 01/21/2026 - 05/05/2026
Days, Time: M, 5:00PM-7:50PM
Location: Davis 113A, North Campus
Credit Hours: 3
Enrollment: 2/30 (2/30 seats reserved for computer science & engineering majors only) (Active)
Info:
CSE 610 Human–Computer Interaction: Foundations for Design and Research (Lecture)
Section: XILU
Instructor: Xi Lu
Description: This graduate-level course provides an introduction to Human–Computer Interaction (HCI), combining practical design methods with an overview of core research in the field. We will cover user needs and task analysis, user interface design guidelines, and methods for prototype evaluation and interface testing, while also examining major research topics and forms of contribution in HCI. Over the past few decades, HCI has expanded from improving computer use in traditional workplaces to investigating and critiquing the many technologies embedded in everyday life. This course will help you understand and engage with that evolution. The course is designed for Master’s and PhD students who may be new to HCI but are interested in practically applying HCI skills and methods: whether to design and evaluate technology or to conduct HCI research.
Instruction Mode: In person
Class #: 23022
Dates: 01/21/2026 - 05/05/2026
Days, Time: M, 11:00AM-1:50PM
Location: Clemen 215, North Campus
Credit Hours: 3
Enrollment: 7/10 (Active)
Info:
CSE 610 Specifical Topic on Mobile Computing: Systems, Networks, Sensing and Mobile AI (Lecture)
Section: YAX
Instructor: Yaxiong Xie
Description: This Special Topics course explores the rapidly evolving world of Mobile Computing, where mobile systems, wireless networks, sensing, and AI come together to power the apps and devices we use every day. We will dive into how smartphones and wearables work at the system level, how they maintain fast and reliable connectivity through Wi-Fi and 5G, and how sensors enable rich applications such as indoor localization, motion tracking, and mobile health monitoring. The course will also cover how to run advanced mobile AI on resource-constrained devices, turning phones into intelligent, real-time decision makers. Through hands-on labs, real mobile measurements, and project-based learning, students will gain practical experience building next-generation mobile applications, understanding how the mobile Internet truly works, and exploring the cutting-edge technologies shaping the future of ubiquitous computing.
Instruction Mode: In person
Class #: 23887
Dates: 01/21/2026 - 05/05/2026
Days, Time: M, 2:00PM-4:50PM
Location: Davis 338A, North Campus
Credit Hours: 3
Enrollment: 0/30 (0/30 seats reserved for computer science & engineering majors only) (Active)
Info:
CSE 703 Advanced Computational Vision (Seminar)
Section: YUAN
Instructor: Junsong Yuan
Description: Topics will cover state-of-the-art computer vision techniques such as image modeling, vision-language alignment, multi-modality model, diffusion model, and vision transformer, as well as their applications in video analytics, virtual/augmented reality, image/video generation, autonomous driving.
Prereqs: CSE 473/573 Computer Vision and Image Processing
Coreqs: Deep learning
Instruction Mode: In person
Class #: 21197
Dates: 01/21/2026 - 05/05/2026
Days, Time: T, 2:00PM-4:50PM
Location: Obrian 108, North Campus
Credit Hours: 1-3
Enrollment: 16/30 (Active)
Info:
CSE 705 Smart Environments – Systems, architecture and applications (Seminar)
Section: SRID
Instructor: Ramalingam Sridhar
Description: CSE-705: Smart Environments – Systems, architecture and applications Spring 2026 (1-3 credits). Time: Tuesday 500PM – 7:00PM Office hours: Mon, Tues 4:00pm - 5:00pm (or by appointment) Instructor: Prof Ramalingam Sridhar 338K Davis Hall rsridhar@buffalo.edu Overview: This graduate-level seminar will cover hardware, systems, security and applications aspects of Smart environments and is intended for graduate students who have taken one or more core courses in CSE. Smart environments are enabled through devices such as Internet of Things making a connected world possible in wide-ranging forms and applications. Concepts addressed will include architecture, networks and security requirements, opportunities in the building & deployment of smart environments and wide-ranging applications. Some of the key challenges in each of these aspects will be presented and possible solutions will be considered. This seminar is for students with background in any of the core areas in CSE, since all can contribute to different aspect of smart environments. In particular, project emphasis will be given to applications in medical, automobile and other niche areas. Requirements: A comprehensive project and a research paper will be expected along with a detailed presentation. The topic addressed in smart environments and applications can align with the student’s background. For further details, please contact Prof Sridhar at rsridhar@buffalo.edu. Grading (S/U): Class attendance and participation is required Project and presentation are required
Notes: This seminar is for students with background in any of the core areas in CSE, since all can contribute to different aspect of smart environments.
URL: https://cse.buffalo.edu/~rsridhar/cse705/
Instruction Mode: In person
Class #: 21150
Dates: 01/21/2026 - 05/05/2026
Days, Time: T, 5:00PM-7:00PM
Location: Davis 113A, North Campus
Credit Hours: 1-3
Enrollment: 5/30 (Active)
Info:
CSE 706 Advanced Topics in Generative AI: Diffusion Models, Personalization, and Reliability (Seminar)
Section: RATH
Instructor: Nalini Ratha
Description: Generative AI has ushered in a new era of creativity and data modeling, where systems can synthesize realistic, diverse, and high-fidelity samples by learning the intricate patterns underlying real-world data distributions. At the core of this revolution lie diffusion models and their continuous-time counterparts, flow-matching models - frameworks that first corrupt data with noise (the forward process) and then learn to reverse that corruption (the backward process). In this seminar, we will trace the evolution of this framework along five key dimensions through extensive paper-reading: 1. Trustworthy Generation — As diffusion models enter safety-critical domains (e.g., medical imaging, autonomous systems, or education), ensuring trustworthiness becomes paramount. We will examine how calibration, interpretability, privacy, and fairness are encoded or neglected within diffusion architectures. Papers on uncertainty quantification, fairness-aware training, and adversarial robustness will guide discussions on how to make generative systems reliable, auditable, and ethically aligned. 2. Personalized Diffusion — Real-world generative systems must adapt to individual preferences and data signatures. We will study mechanisms for personalization and fine-grained control: from user-conditioned denoisers and LoRA-based adapters to latent-space alignment and continual personalization without forgetting. These methods raise key questions about balancing adaptability with data privacy and generalization. 3. Efficient Sampling and Training — While diffusion and flow-matching models achieve stunning fidelity, they are often computationally expensive. We will cover efficiency-focused works such as accelerated samplers, consistency models, distillation, rectified flows, and score distillation sampling. Discussions will emphasize theoretical trade-offs between accuracy, stability, and cost - paving the way toward lightweight, deployable diffusion systems. 4. Faithful and Grounded Generation (No Hallucination) — Hallucination remains a persistent weakness of generative models. We will explore techniques for factual grounding, retrieval-augmented conditioning, and energy-based regularization to ensure generated outputs remain consistent with real data or evidence. A key goal is to understand the boundary between creativity and correctness, and how architectural or training constraints can enforce semantic fidelity. 5. Mode Diversity and Collapse Prevention — Despite their expressive power, diffusion models can still suffer from mode collapse, favoring frequent patterns while underrepresenting rare or diverse structures. We will review theoretical perspectives linking this issue to score approximation and data imbalance, and investigate remedies such as entropy-based objectives, guidance diversity regularization, and optimal transport (OT)–based balancing. The aim is to cultivate a nuanced understanding of diversity preservation in generative distributions. Through guided readings of research papers from top-tier conferences and journals, including NeurIPS, ICML, ICLR, TPAMI, participants will gain mathematical intuition, critical analysis skills, and research-level fluency in designing trustworthy, personalized, efficient, and diversity-preserving generative models capable of producing faithful and high-quality outputs across domains.
Prereqs: Deep Learning, Computer Vision
Instruction Mode: In person
Class #: 19342
Dates: 01/21/2026 - 05/05/2026
Days, Time: W, 3:00PM-5:50PM
Location: Talbrt 115, North Campus
Credit Hours: 1-3
Enrollment: 13/20 (Active)
Info:
CSE 711 AI-Driven Wireless Sensing and Next-Generation Communication Systems (Seminar)
Section: AYY
Instructor: Sai Roshan Ayyalasomayajula
Description: This seminar explores the cutting edge of wireless systems at the intersection of signal processing, machine learning, and next-generation communication technologies. Students will learn how AI can both interpret and generate wireless signals — from device-free human activity recognition using WiFi and mmWave, to neural-radiance RF fields that model propagation as differentiable digital twins, to deep-learning-enabled integrated sensing and communication (ISAC) for adaptive 6G networks. The course blends theory, hands-on MATLAB/ML labs, and project-based learning, giving students experience reproducing state-of-the-art systems such as SLNet, SenseFi, NeRF², NeWRF, and RF-Genesis. Through weekly lectures, paper discussions, coding assignments, and a semester-long project, students will gain a holistic understanding of how AI and ML are reshaping wireless sensing and communication, equipping them to design, implement, and critically evaluate next-generation wireless algorithms.
Notes: Some more Prerequisites: Programming concepts: Python and PyTorch/Tensorflow Optional: Background in signal processing, optimization, and machine learning may allow you to better appreciate certain aspects of the course material.
Prereqs: Linear Algebra, Intro to ML, Computer Vision, Probability and Statistics
Instruction Mode: In person
Class #: 21149
Dates: 01/21/2026 - 05/05/2026
Days, Time: W, 1:00PM-3:50PM
Location: Davis 338A, North Campus
Credit Hours: 1-3
Enrollment: 7/30 (Active)
Info:
CSE 712 Introduction to AI for Health (Seminar)
Section: WEN
Instructor: Wenyao Xu
Description: Introduction to AI for Health is a graduate-level seminar that explores emerging research, methods, and applications of artificial intelligence in healthcare, medicine, and biomedical engineering. The course is designed to give students both a broad conceptual foundation and a deep, research-oriented understanding of how modern AI techniques are transforming health diagnostics, therapeutics, monitoring, rehabilitation, and clinical decision-making. This is a discussion-driven seminar rather than a lecture-based class. Each week, students will engage with cutting-edge research papers from top conferences and journals (e.g., EMBS, ICHI, NeurIPS, ICML, MICCAI, Nature Medicine). For each topic area, the instructor will curate two or more key readings, and students will select additional papers of interest to present. Presentations will focus on problem motivation, methodological contributions, experimental design, and implications for real-world deployment in healthcare settings. In addition to reading and presenting research, students may optionally pursue a course project, individually or in teams. Projects may involve building prototypes, evaluating datasets, reproducing existing models, conducting error analyses, or designing new AI-for-health solutions. Projects are intended to help students exercise creativity and research skills, especially those who wish to explore new ideas or prepare for future thesis work. Topics may include (but are not limited to): - AI in medical imaging and diagnostics - Sensor-based health monitoring and wearable AI - Multimodal learning for health - AI-driven rehabilitation and assistive technologies - Bio-signal processing (EEG, EMG, ECG) with machine learning - AI for population health, epidemiology, and public health policy - Trustworthy, explainable, and ethical AI in healthcare - Regulatory, safety, and translational pathways for AI medical technologies By the end of the course, students will gain: * A broad and current understanding of the AI-for-health research landscape * Experience critically reading and discussing scientific papers * Presentation and scientific communication skills * (Optional) Hands-on experience creating or analyzing AI-based health solutions This seminar is ideal for graduate students in computer science, engineering, robotics, data science, biomedical engineering, and related health-tech disciplines who want to understand the research frontier of AI-driven healthcare innovation.
Notes: If you’re interested in how AI can improve healthcare, this seminar is a great place to start. You’ll read exciting research papers, present your ideas, and even try a project of your own if you choose. No prior health background required—just curiosity and a willingness to learn.
Prereqs: None
Instruction Mode: In person
Class #: 23886
Dates: 01/21/2026 - 05/05/2026
Days, Time: T, 2:00PM-4:50PM
Location: Davis 113A, North Campus
Credit Hours: 1-3
Enrollment: 5/30 (5/30 seats reserved for computer science & engineering majors only) (Active)
Info:
CSE 713 Software Analysis and Applications (Seminar)
Section: HAI
Instructor: Haipeng Cai
Description: Modern software systems—from web services and mobile platforms to distributed microservices and AI-enabled applications—demand rigorous methods to ensure security, reliability, privacy, and performance. This seminar explores the cutting-edge foundations, techniques, and emerging research directions in software analysis: static, dynamic, hybrid, and AI-augmented approaches. Students will engage deeply with seminal papers and state-of-the-art research. The course is intentionally broad: we study not only program analysis but also system-level, cross-language, and AI-integrated software analysis spanning the full lifecycle of software artifacts. Special emphasis is given to security and privacy applications, but we also investigate analysis for functional correctness, robustness, and performance optimization. Students will learn both classical analysis techniques and how modern advancements (e.g., LLM-based analysis, agentic AI systems, hybrid symbolic-neural analysis, fuzzing with AI guidance) are reshaping software analysis research and practice. =========== Learning Objectives ============== By the end of the course, students will: - Understand foundational and advanced techniques in software analysis, including their theoretical underpinnings and practical tradeoffs. - Analyze and critique research papers, with emphasis on rigor, novelty, and empirical evaluation. - Gain hands-on experience with modern analysis tools (e.g., static analyzers, fuzzers, symbolic executors, LLM-based analyzers). - Connect analysis techniques to impactful applications in software security, privacy, reliability, and performance engineering. - Develop and present a mini research project exploring an emerging direction in the field. =========== Main Topics & Modules =========== Below is a representative outline (topics may be adjusted to align with student interest and cutting-edge developments): 1. Foundations of Software and Program Analysis 2. Modern Software Analysis for Security & Privacy 3. Runtime Support and Techniques 4. AI-Enhanced and Data-Driven Software Analysis 5. Software Analysis for Performance, Reliability, and Testing =========== Format & Expectations =========== This is a research-oriented seminar, emphasizing: 1. paper presentation & discussion. Each student presents several research papers, leading discussion and critique. Papers will include classics as well as the most recent advances. 2. Mini-Project (Recommended for PhD Students) Students may explore new ideas, replicate a recent result, or build a proof-of-concept analysis tool. Projects can be aligned with ongoing research for publication-quality results. 3. Participation Active participation in discussions is required. Students are expected to read each paper carefully and come prepared with insights and questions. ******************** Why Take This Course? ******************** - Front-line relevance: Software analysis is central to modern security, privacy, reliability, and performance engineering—industries are hungry for these skills. - Research value: The seminar reflects the latest research trends, preparing students for top-tier conference publications and dissertation topics. - Cutting-edge content: We cover both foundational principles and emerging AI-driven techniques reshaping program analysis today. - Hands-on exposure: Students learn real tools used in academia and industry: fuzzers, symbolic executors, static analyzers, LLM-based agents, etc. - Interdisciplinary impact: Skills learned here benefit research in software engineering, cybersecurity, systems, and AI. This seminar is aimed at students who want to deeply understand how modern software is analyzed—and how software analysis is evolving with the rise of AI.
Instruction Mode: In person
Class #: 24078
Dates: 01/21/2026 - 05/05/2026
Days, Time: M, 2:30PM-3:50PM
Location: Talbrt 111, North Campus
Credit Hours: 1-3
Enrollment: 1/30 (Active)
Info:
CSE 750 Databases and Programming Languages (Seminar)
Section: DBPL
Instructor: Qianchuan Ye
Description: This seminar will be co-located with the weekly meeting of the databases and programming languages group, which consists of weekly talks and roundtable discussions. Students will be graded on (i) Attendance and (ii) Presentations. Specifically, (i) Students are expected to attend all weekly meetings; and (ii) Students are expected to present at least once, and more if taking more credit hours. The presentation schedule will be determined the first week of the seminar. Typically the talks will be 50 minutes, including Q&A. Topics of the talks include students' own research, seminal papers, and state-of-the-art in the area of databases and/or programming languages.
Instruction Mode: In person
Class #: 23876
Dates: 01/21/2026 - 05/05/2026
Days, Time: T, 12:30PM-2:00PM
Location: Davis 310, North Campus
Credit Hours: 1-3
Enrollment: 1/12 (0/12 seats reserved: force registration only) (Active)
Info:
CSE 750 Theory (Seminar)
Section: THEO
Instructor: Atri Rudra
Description: This seminar will be co-located with the weekly meeting of the theory group, which consists of weekly talks. Students will be graded on (i) Attendance and (ii) Presentations. Specifically, (i) Students are expected to attend all weekly meetings; and (ii) Students are expected to present two times the number of credits they sign up for. So 1 credit, a student will present twice, for 2 credits four times and for 3 credits 6 times. The exact kind of presentations (e.g. which set of paper/notes as well as the length of each presentation) will be determined the first week of the seminar (based on the interest of the attendees as well as the number of registered students).
Instruction Mode: In person
Class #: 23877
Dates: 01/21/2026 - 05/05/2026
Days, Time: W, 12:00PM-2:00PM
Location: Davis 310, North Campus
Credit Hours: 1-3
Enrollment: 0/12 (0/12 seats reserved: force registration only) (Active)
Info: