BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.4.12//EN
TZID:Asia/Kolkata
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
UID:229@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20261023T110000
DTEND;TZID=Asia/Kolkata:20261023T120000
DTSTAMP:20261028T070345Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-colloquium-102-cds-23-octobe
 r-2026-collaboration-amid-differences-generalization-adaptation-incentiviz
 ed-participation-and-minimax-optimization-in-federated-learning/
SUMMARY:Ph.D: Thesis Colloquium: 102: CDS: 23\, October 2026 "Collaboration
  amid Differences: Generalization\, Adaptation\, Incentivized Participatio
 n and Minimax Optimization in Federated Learning"
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nPh.D. Thesis Col
 loquium\n\n\n\nSpeaker: Mr. Yashwanth. M\nS.R. Number: 06-18-01-10-12-21-1
 -19272\nTitle: "Collaboration amid Differences: Generalization\, Adaptatio
 n\, Incentivized Participation and Minimax Optimization in Federated Learn
 ing"\nResearch Supervisor: Prof. Anirban Chakraborty\nDate &amp\; Time : O
 ctober 23\, 2026 (Friday)\, 11:00 AM\nVenue : #102\, CDS Seminar Hall\n\n\
 n\nABSTRACT\n\nFederated Learning (FL) is a distributed machine learning p
 aradigm\, where multiple clients train a model collaboratively under the o
 rchestration of a central server while keeping the training data private t
 o respective clients. Since its introduction\, FL has been enabling numero
 us privacy-preserving applications across major industries like healthcare
  and big-tech. Despite its advantages\, the federated framework introduces
  several critical algorithmic challenges under practical constraints. This
  thesis studies federated machine learning tasks under a number of these c
 hallenges\, as described below. \nA key challenge in cross-device and cro
 ss-silo FL is that the data held by individual clients is typically not in
 dependent and identically distributed (non-IID). Under this heterogeneity\
 , each client's model drifts toward its own local minimum\, even when all 
 clients participate in the federation. The global model\, aggregated from 
 these drifted local models\, then performs poorly. A single global model d
 oes have one clear advantage: it can be deployed to clients that were neve
 r seen during training. But because no single model suits diverse datasets
  well\, many methods turn to partial model personalization. Relying on loc
 al parameters\, however\, leaves the server unable to serve new\, unseen c
 lients that never joined training. Heterogeneity thus produces both the 
 ‘client-drift’ problem and a personalization-generalization trade-off.
  A second challenge originates from a practical point that traditional FL 
 tends to ignore: clients are assumed to participate whether or not they be
 nefit. This is often unrealistic\, and encouraging participation would nee
 d a suitable incentive mechanism. FL under client rationality with non-mon
 etary incentives remains unstudied and existing algorithms do little to pr
 event federation collapse in such settings\, especially when data is highl
 y heterogeneous. A third challenge arises from the data locality constrain
 t itself. Because data stays strictly on each client\, objectives like min
 imax optimization introduce variables tied to individual data points that 
 the server cannot aggregate\, leaving their aggregation undefined. This th
 esis takes up each of these challenges in turn\, formulating and addressin
 g practical\, non-trivial FL problems that arise from them. \nFirst\, to 
 address client-drift due to label heterogeneity\, we introduce Adaptive Se
 lf-Distillation (ASD)\, a client-side regularizer that adapts to each clie
 nt's local dataset using the global model's prediction entropy together wi
 th the client’s label distribution. We justify theoretically that ASD re
 duces client drift and show empirically that it improves generalization as
  it results in the global model converging to a 'flat’ minimum. Because 
 ASD is plug-and-play\, integrating it into standard FL methods is straight
 forward\, and across multiple real-world benchmarks it leads to substantia
 l gains. Building on this connection of generalizability to flatness of lo
 ss topology\, we next aim to achieve the generalization objective explicit
 ly through a flatness-constrained optimization framework that controls the
  largest eigenvalue of the Hessian of training-loss. To keep this tractabl
 e during local training\, we propose FedMAN\, a computationally efficient 
 regularizer that minimizes the per-layer activation norm of the client-sid
 e models. We provide theoretical guarantees for  neural nets with fully c
 onnected and convolutional layers\, showing that minimizing this layer-wis
 e activation norm bounds the top Hessian eigenvalue of the client's loss\,
  driving the aggregated model toward a flat minimum. Integrating this regu
 larizer with existing FL methods delivers significant gains in accuracy. 
 \nWe now explore how we can adapt a pre-trained model in a federated setup
 \, with clients training on local datasets that are unlabeled and exhibit 
 significant domain gaps from one client to another. Towards this end\, we 
 address the Federated Source-Free Domain Adaptation (FFreeDA) problem\, wh
 ere clients train locally on such unlabeled training samples from widely h
 eterogeneous data distributions. The client models are initialized with a 
 pre-trained server model trained on a source domain that is significantly 
 different from every client's local data distribution and also unavailable
  during the federated training. Adapting Source-Free Domain Adaptation (SF
 DA) methods to this federated setting struggles with client drift: the ext
 reme heterogeneity from the domain gaps results in unreliable pseudo-label
 s. We introduce FedSCAl\, an FL framework built on our Server-Client Align
 ment (SCAl) mechanism\, which regularizes client updates by matching the c
 lient and server model predictions on weak and strong augmentations of the
  input data\, thereby achieving strong server-client alignment\, reduced c
 lient-drift as well as more reliable pseudo labels to train local models o
 n. Extensive experiments on benchmark vision datasets show that FedSCAl ou
 tperforms the competing FL methods. \n\nIn the works described above\, we
  have trained a global model in which the clients share all model paramete
 rs during the training. However\, to combat data heterogeneity\, partial m
 odel personalization methods have been proposed. The main drawback here is
  that the server model cannot be used for clients which did not participat
 e in the federation. This creates a trade-off in balancing personalization
  and generalization. In the context of Vision transformer models\, when th
 e training paradigm is prompt tuning\, we propose PEP-FedPT\, a framework 
 achieving both generalization and personalization in federated prompt tuni
 ng of ViTs. At its core is the Class-Contextualized Mixed Prompt (CCMP) - 
 class-specific prompts maintained alongside a globally shared prompt\, whi
 ch CCMP adaptively combines per input using weights from global class prot
 otypes and client class priors. This gives per-sample prompt personalizati
 on without storing any client-dependent trainable parameters and the promp
 ts are optimized collaboratively via federated averaging. We demonstrate t
 he efficiency of our method on  CIFAR-100\, TinyImageNet\, DomainNet and 
 iNaturalist datasets. \n\nSo far\, we have assumed that clients take part
  in the federation willingly under server orchestration  and without cond
 itions. In practice\, this is often not the case. Clients are usually self
 -interested and a client may care more about how well the shared global mo
 del performs on its held-out data relative to its own local model performa
 nce or any desired threshold. If the clients do not see the benefit they d
 rop out and the whole federation can fall apart. One way to keep the clien
 ts engaged is to pay them\, and such monetary incentives have been studied
  before. In this work we take a different approach with FedUCA\, a non-mon
 etary framework built on a non-payment channel between server and clients.
  The server first shares the exact aggregation weights with each client\, 
 from which the client designs a utility and signals how willing it is to p
 articipate. The server then chooses aggregation parameters satisfying each
  client's individual-rationality (IR) constraint\, sustaining global model
  performance. Extensive experiments on standard datasets show that\, by pr
 ioritizing feasibility of participation\, FedUCA achieves higher client re
 tention and\, in turn\, a stronger global model. As an extension\, we desi
 gn FedLoPers\, a lottery-based persuasion scheme that randomizes the infor
 mation shared with clients. Rather than committing to the true aggregation
  weights\, the server shares a distribution over the same. This lottery-in
 spired design further improves client participation. \n\nIn all of the ab
 ove works\, the clients minimize their respective loss functions. We now c
 onsider the federated minimax problems\, a class of problems that includes
  distributionally robust optimization and adversarial training. Such probl
 ems carry an inner maximization variable that is a worst-case reweighting 
 of a client's own samples or a perturbation of its own inputs\, constructe
 d entirely from local data. Averaging such variables across clients is not
  just suboptimal but ill-posed. We therefore study the federated minimax p
 roblem in which the minimization variable is shared and aggregated while e
 ach client retains a personalized maximization variable that is never shar
 ed. In the nonconvex-concave regime\, our proposed projected personalized 
 ascent scheme attains the standard sublinear rate. In the nonconvex-PL reg
 ime\, the same scheme incurs a stepsize-invariant noise floor from the sto
 chastic ascent updates. To remove it\, we propose VR-SGDA\, a variance-red
 uced ascent step. This eliminates the floor and attains the sublinear rate
 . Experiments on subpopulation-robustness and adversarial training confirm
  our theoretical findings and contributions. \n\n\n\nALL ARE WELCOME
CATEGORIES:Events,Ph.D. Thesis Colloquium
END:VEVENT
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
X-LIC-LOCATION:Asia/Kolkata
BEGIN:STANDARD
DTSTART:20251023T110000
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR