BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//6.4.6//EN
TZID:Asia/Kolkata
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
UID:16@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata;VALUE=DATE:20231117
DTEND;TZID=Asia/Kolkata;VALUE=DATE:20231119
DTSTAMP:20231102T051946Z
URL:https://cds.iisc.ac.in/events/workshop-on-tensor-computation-and-machi
ne-learning-tcml-november-17-18-2023/
SUMMARY:Workshop on Tensor Computation and Machine Learning (TCML)\, Novemb
er 17-18\, 2023.
DESCRIPTION:Click here for program details.
CATEGORIES:Events
END:VEVENT
BEGIN:VEVENT
UID:15@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20231102T140000
DTEND;TZID=Asia/Kolkata:20231102T150000
DTSTAMP:20231102T051759Z
URL:https://cds.iisc.ac.in/events/m-tech-research-thesis-defense-cds-02-no
vember-2023%e2%80%b3semi-analytical-solution-for-eigenvalue-problems-of-la
ttice-models-with-boundary-conditions/
SUMMARY:M.Tech Research: Thesis Defense: CDS: 02\, November 2023″Semi-ana
lytical solution for eigenvalue problems of lattice models with boundary c
onditions”.
DESCRIPTION:02 Nov @ 2:00 PM -- 3:00 PM\n\nDEPARTMENT OF COMPUTATIONAL AND
DATA SCIENCES\nM.Tech Research Thesis Defense\n\n_________________________
_________________________________________________________________\n\nSpeak
er : Ms. Athira Gopal\n\nS.R. Number : 06-18-00-10-12-19-2-17782\n\nTitle
: “Semi-analytical solution for eigenvalue problems of lattice models wi
th boundary conditions”\n\nResearch Supervisor: Prof. Murugesan Venkatap
athi\n\nDate &\; Time : November 02\, 2023 (Thursday)\, 02:00 PM\n\nVen
ue : Room No. 102 (CDS Seminar Hall)\n\n__________________________________
________________________________________________________\nAbstract\nClosed
-form relations for limiting eigenvalues of an infinite k-periodic spatial
lattice in any number of dimensions ‘d’\, and its semi-analytical ext
ensions for any given size ‘n’ of the lattice with free-free boundary
conditions\, are known. These are based on the eigenvalues of tridiagonal
k-Toeplitz matrices (representing chains and d=1)\, and their tensor produ
cts or sums. These semi-analytical methods for eigenvalues incur drastical
ly lower computing costs than the direct numerical methods i.e. O(n) vs. O
(n^2) for the latter\, and further they are more accurate for sufficiently
large lattices approaching the limiting case (n >\; 100). This advantag
e in computing cost\, accuracy\, and numerical stability results as the or
iginal eigenvalue problem of nk in size is reduced to n eigenvalue problem
s each k in size\, further making this approach very amenable to parallel
computation when required. In this work\, their errors in eigenvalues are
compared with the errors of the direct numerical methods using special exa
mples with high condition numbers. Secondly\, in the absence of such analy
tical methods\, one also resorts to periodic boundary conditions to limit
the size of the numerical model representing a very large system. The conv
ergence of numerical models with periodic boundary conditions to the limit
ing eigenvalues is highlighted\, to emphasize the utility of the closed-fo
rm solution for the limiting eigenvalues. Thirdly\, the fixed-fixed bounda
ry conditions on a finite chain and their counterpart for periodic spatial
lattices in higher dimensions (d>\;1) are addressed using perturbations
to tridiagonal k-Toeplitz matrices on their main diagonal. Extensions of
the semi-analytical methods for these cases by applying numerical methods
to update only the few perturbed eigenvalues is proposed. An efficient ext
ension for evaluating the eigenvectors in the case of real eigenvalues as
required in most physical systems\, is also presented.\n\n================
===============================================================\n\nALL ARE
WELCOME
CATEGORIES:Events,Thesis Defense
END:VEVENT
BEGIN:VEVENT
UID:14@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230920T100000
DTEND;TZID=Asia/Kolkata:20230920T110000
DTSTAMP:20231102T051640Z
URL:https://cds.iisc.ac.in/events/seminar-cds-102-20th-september-applicati
ons-of-hahn-banach-theorem-to-sequence-spaces-and-variational-inequality/
SUMMARY:{Seminar} @ CDS: #102 : 20th September: “Applications of Hahn- Ba
nach Theorem to Sequence Spaces and Variational Inequality.”
DESCRIPTION:20 Sep @ 10:00 AM -- 11:00 AM\n\nDepartment of Computational an
d Data Sciences\n\nDepartment Seminar\n\n \;\n\nSPEAKER : Sudarsan Nan
da (Retd. Professor- IIT Kharagpur) is at present the Professor of Eminenc
e and Research Chair at KIIIT\, Deemed to be University\, Bhubaneswar.\n\n
TITLE : “Applications of Hahn- Banach Theorem to Sequence Spaces and Var
iational Inequality.”\n\nDate &\; Time : September 20\, 2023\, 10:00
AM\n\nVenue : # 102\, CDS Seminar Hall\n\n================================
===========================================================\nAbstract\nApp
lication of the Hahn-Banach Theorem to the space of bounded sequences with
a specific sublinear functional p defined on it gives rise to linear func
tionals\, which are dominated by p and are extensions of limits of converg
ent sequences. These are called Banach Limits\, which were studied by Bana
ch (1932)\, and their uniqueness is called almost convergence and was char
acterized by Lonentz [1948].\n\nIn the present lecture\, we will discuss a
bout the absolute analog of almost convergence\, which generalizes LP spac
es. The two concepts Variational Inequality and complementarily problems a
re essentially the same concepts which are studied by two different groups
of mathematicians: applied mathematics on one hand and operations researc
hers on the other hand. The proof existence of the variational inequality
problem uses Hanh- Banach Theorem or Fixed-Point theorems. In this lecture
\, we will discuss about the existence of solutions of the complementary p
roblem\, under the most general conditions on the operator and the cone.\n
Biography\nSudarsan Nanda (Retd. Professor- IIT Kharagpur) is at present t
he Professor of Eminence and Research Chair at KIIIT\, Deemed to be Univer
sity\, Bhubaneswar. He is a Ph.D. and D.Sc. in Mathematics and was a Profe
ssor at IIT Kharagpur. He was Vice-Chancellor\, North Orissa University. P
rofessor Nanda was a Visiting Professor at the Universities of Guelph\, Ca
nada\, Pisa\, and Milan in Italy\, University of Central Florida\, USA and
Chinese University of Hongkong. He was an Associate member of ICTP\, Trie
ste\, Italy\, visited University of Kaiserslautern\, Germany under exchang
e programme and several Universities of Europe and USA. S. Nanda is a Fell
ow of Institute of Mathematics and Applications\, U.K. (FIMA)\, Forum D’
e Analysts\, Member of the Editorial Board of the Journal of Fuzzy Mathema
tics and several other Journals on Mathematics. He is a reviewer of Mathem
atical reviews and several journals of international repute. Prof. Nanda r
eceived the distinguished teacher award during 2002 and received Fulbright
and AMS travel grant for visiting USA\, during 2000. He has received dist
inguished senior scientist award from Odisha Bigyan Academy\, Bhubaneswar
Chapter of Indian Science Congress association\, Odisha environment societ
y . S. Nanda has published around 235 research papers in the journals of I
nternational repute\, guided 22 Ph. D students and authored and co-authore
d 20 books and many popular articles relating to Mathematics and Higher ed
ucation.\n\nTaking an active interest in the modernization of the Mathemat
ics syllabus\, established research culture and popularization of mathemat
ics\, which has a great impact on society S. Nanda’s research work has b
een widely cited and he has been a member in many academic bodies. He has
done social work by establishing a school for mentally handicapped childre
n at IIT Kharagpur Campus. As Vice Chancellor of North Orissa University i
ntroduced Santali language as MIL for Under Graduate students.\n\nHost Fac
ulty: Dr. Ratikanta Behera\n==============================================
====================================\nALL ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:13@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230913T100000
DTEND;TZID=Asia/Kolkata:20230913T110000
DTSTAMP:20231102T051511Z
URL:https://cds.iisc.ac.in/events/seminar-cds-102-13th-september-simulatio
n-data-6-steps-from-theory-to-impact-through-disasters-engineering-heritag
e/
SUMMARY:{Seminar} @ CDS: #102 : 13th September: “Simulation & Data: 6 Ste
ps from Theory to Impact through Disasters\, Engineering & Heritage.”
DESCRIPTION:Department of Computational and Data Sciences\n\nDepartment Sem
inar\n\n \;\n\nSPEAKER : Dr. Phil Weir\, Founder and Director of flax
&\; Teal Ltd\, UK.\n\nTITLE : “Simulation &\; Data: 6 Steps from T
heory to Impact through Disasters\, Engineering &\; Heritage.”\n\nDat
e &\; Time : September 13\, 2023\, 10:00 AM\n\nVenue : # 102\, CDS Semi
nar Hall\n\n==============================================================
=============================\nAbstract\nGeospatial data\, open data\, IoT
\, computational physics and streaming infrastructure metrics – working
across varied areas\, we see the simplicity of standard approaches struggl
ing to translate to complex\, messy\, multi-domain problems. In industry\,
theoretical challenges combine in nonlinear ways and success metrics are
rarely neatly defined. The focus on accessible\, reusable\, scalable data
solutions in short timescales can be a huge hurdle to newcomers used to re
fined\, well-understood patterns. Moreover\, as the capability of software
tooling and hardware surges\, a static skillset quickly becomes irrelevan
t. This presentation will outline six key insights about industrial mathem
atics\, to help surf that wave.\nBiography\nDr. Phil has his Bachelor’s
degree and Master’s Degree in Pure Mathematics from the University of Ca
mbridge. He completed his PhD in Computational Mathematics from the Univer
sity of Otago\, New Zealand in 2012. He also worked as Web Developer at va
rious companies in the UK. He worked as a Software Development Engineer at
NUMA Engineering Services Ltd\, Dundalk\, Ireland from 2013-2016. He is a
Board Member of OpenUK. He is the Founding Director and Executive Directo
r of Flax &\; Teal Limited\, Belfast\, UK. His areas of expertise inclu
de Data Science\, Scientific Computing\, Numerical Analysis\, Web-based Si
mulation\, and On-Demand Data Analysis. He also has work experience with g
eospatial data. Please visit his company profile here: https://flaxandteal
.co.uk/pages/team/ and his profile here https://www.linkedin.com/in/phil-w
eir-033b5a62/ His company works on Scientific Computing and Data Science.\
n\nHost Faculty: Dr. Ratikanta Behera\n===================================
===============================================\nALL ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:12@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230911T100000
DTEND;TZID=Asia/Kolkata:20230911T110000
DTSTAMP:20231102T051329Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-colloquium-cds-unsupervised-
test-time-adaptation-for-patient-specific-deep-learning-models-in-medical-
imaging/
SUMMARY:Ph.D. Thesis {Colloquium}: CDS : “Unsupervised test-time adaptati
on for patient-specific deep learning models in medical imaging.”
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nPh.D. Thesis Col
loquium\n\n_______________________________________________________________
___________________________\n\nSpeaker : Mr. Hariharan Ravishankar\n\nS.R.
Number : 06-18-00-11-12-20-1-18917\n\nTitle : “Unsupervised test-time a
daptation for patient-specific deep learning models in medical imaging”\
n\nResearch Supervisor: Prof. Phaneendra Kumar Yalavarthy\n\nDate &\; T
ime : September 11\, 2023 (Monday) at 10:00 AM\n\nVenue : # 102 CDS Semina
r Hall\n\n________________________________________________________________
__________________________\nAbstract\nDeep learning (DL) models have achie
ved state-of-the-art results in multiple medical imaging applications\, re
sulting in the widespread adoption of artificial intelligence (AI) models
for radiological workflows. Despite their success\, they exhibit two major
weaknesses in real-world applications: a) poor generalization and b) lack
of patient optimality. The potential “distribution shift” in unseen m
edical imaging data\, owing to changes in demography\, acquisition hardwar
e\, etc.\, often leads to a reduction in performance. Despite the high ave
rage performance\, they are prone to failures on “individual” cases wi
th minor input modulations than the training data. Solving this problem is
critical in medical image analysis because this variance in performance a
cross cases and abject failures on select subjects will increase the burde
n on care-giving experts and reduce trust in AI-based applications.\n\nThe
aim of this thesis work is to develop methods for patient-specific test-t
ime adaptation of pre-trained DL models on individual subject data. While
generalization is often treated as a training time goal\, test-time adapta
tion (TTA) modifies the weights during inference time in an unsupervised m
anner specific to each subject data to jointly address generalization and
patient optimality. While sustaining performance in individual subjects is
seen as a challenge for current DL models\, this line of research also ha
s an opportunity for next-generation clinical decision support systems. AI
methods that would result in patient-specific DL models can be heralded a
s the next big step in medical imaging.\n\nInformation Geometric Test Time
Adaptation (IGTTA) for semantic segmentation: The test-time adaptation (T
TA) of deep-learning-based semantic segmentation models\, specific to indi
vidual patient data\, was addressed in this part of thesis work. Existing
TTA methods in medical imaging are often unconstrained and require prior a
natomical information or additional neural networks built during the train
ing phase\, making them less practical and prone to performance deteriorat
ion. In this part of thesis work\, a novel framework based on information
geometric principles was proposed to achieve generic\, off-the-shelf\, reg
ularized patient-specific adaptation of models during test-time. By consid
ering the pre-trained model and adapted models as part of statistical neur
omanifolds\, test-time adaptation was treated as constrained functional re
gularization using information geometric measures\, leading to improved ge
neralization and patient optimality. The efficacy of the proposed approach
was shown on three challenging problems: a) improving the generalization
of state-of-the-art models for segmenting COVID-19 anomalies in Computed T
omography (CT) images\, b) cross-institutional brain tumor segmentation fr
om magnetic resonance (MR) images\, and c) segmentation of retinal layers
in Optical Coherence Tomography (OCT) images. Furthermore\, it was demonst
rated that robust patient-specific adaptation can be achieved without addi
ng a significant computational burden\, making it the first of its kind ba
sed on information geometric principles.\n\nTest-time adaptation via domai
n transforms (TTADT): OCT imaging has emerged as the modality of choice fo
r the diagnosis of retinal diseases. The popular\, cost-effective variant
of OCT – Spectral-Domain Optical Coherence Tomography (SDOCT) often suff
ers in image quality owing to speckle noise\, making expert or automated a
nalysis of these images challenging. This part of thesis work proposes for
the first-time inference time adaptation of deep learning models for a gi
ven test sample to achieve improved diagnostic performance. The adaptation
is driven by domain transforms – a framework in which domain-specific a
ugmentation is employed for a test sample\, and the deep learning model we
ights are updated in a constrained and regularized manner. Performance of
the proposed approach was evaluated on both simulated and real-world noisy
B-mode OCT images affected by varying degrees of speckle noise from four
benchmarking SD-OCT datasets. Systematic studies in this part of thesis wo
rk clearly demonstrate the utility of the proposed approach in building ge
neralized and robust deep learning models for automated retinal disease di
agnosis using OCT images.\n\nUncertainty-aware test-time adaptation for in
verse problems (UATTA): Deep learning models for inverse problems (both on
e-pass and unrolled versions): have these limitations 1) inability to prov
ide pixel-wise uncertainty quantification 2) sensitivity to variation in f
orward model and 3) sub-optimal regularization over different iterations.
In this part of thesis work\, uncertainty aware test-time adaptation of of
f-the-shelf models is proposed for improved image reconstruction performan
ce. Different pixel-wise uncertainty quantification methods were explored
for utility in test-time adaptation\, along with data fidelity constraints
. The results were studied for two relevant problems:1) accelerated MRI ac
quisition and 2) quantitative susceptibility mapping for quantifying tissu
e magnetic susceptibility. The proposed framework provides a reliable and
accurate reconstruction of these images.\n\nTest-time adaptation with foun
dational models as neural surrogates (TTANS): Recently\, foundational mode
ls have demonstrated impressive promptable segmentation of natural images.
However\, these models often perform poorly with medical imaging data. In
this part of the thesis\, a two-way synergic approach for jointly utilizi
ng foundational models while adapting pretrained medical imaging segmentat
ion models is explored. The results are demonstrated on various medical im
aging semantic segmentation tasks to show the ability of foundational mode
ls as neural surrogates\, thus demonstrating test-time adaptation achieved
in real time. As these foundational models are capable of handling high l
evels of semantics\, using them as surrogates helped reduce reliance on co
stly annotated data needed to achieve state-of-the-art results.\n\nIn summ
ary\, this thesis work developed methods to improve the generalization of
typical deep-learning models utilized in medical imaging by personalizing
them to individual patients during testing in an unsupervised and automate
d manner. The developed methods demonstrated the applicability and utility
in different medical imaging tasks across various medical imaging modalit
ies. This thesis work is the first of its kind in medical imaging to showc
ase the utility of test-time adaptation methods\, paving the way for furth
er research towards the goal of precision medicine.\n\n===================
============================================================\n\nALL ARE WE
LCOME
CATEGORIES:Events,Ph.D. Thesis Colloquium
END:VEVENT
BEGIN:VEVENT
UID:11@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230830T110000
DTEND;TZID=Asia/Kolkata:20230830T120000
DTSTAMP:20231102T051101Z
URL:https://cds.iisc.ac.in/events/seminar-cds-102-30th-august-recon-all-cl
inical-cortical-analysis-of-heterogeneous-clinical-brain-mri-scans/
SUMMARY:{Seminar} @ CDS: #102 : 30th August: “Recon-all-clinical Cortical
analysis of heterogeneous clinical brain MRI scans”
DESCRIPTION:Department of Computational and Data Sciences\nDepartment Semin
ar\n\nSPEAKER : Karthik Gopinath is a Postdoctoral Research Fellow at Athi
noula A.\n\nTITLE : “Recon-all-clinical Cortical analysis of heterogeneo
us clinical brain MRI scans“\n\nDate &\; Time : August 30\, 2023\, 11
:00 AM\n\nVenue : # 102\, CDS Seminar Hall\n\n============================
===============================================================\nAbstract\
nSurface analysis of the cortex is ubiquitous in human neuroimaging with M
RI\, e.g.\, for cortical registration\, parcellation\, or thickness estima
tion. The convoluted cortical geometry requires isotropic scans (e.g.\, 1m
m MPRAGEs) and good gray-white matter contrast for 3D reconstruction. This
precludes the analysis of most brain MRI scans ac- quired for clinical pu
rposes. Analyzing such scans would enable neuroimaging studies with sample
sizes that cannot be achieved with current research datasets\, particular
ly for underrepresented populations and rare diseases. Here we present the
first method for cortical reconstruction\, registration\, parcellation\,
and thickness estimation for clinical brain MRI scans of any resolution an
d pulse sequence. The methods have a learning component and a classical op
timization module. The former uses domain randomization to train a CNN tha
t predicts an implicit representation of the white matter and pial surface
s (a signed distance function) at 1mm isotropic resolution\, independently
of the pulse sequence and resolution of the input. The latter uses geomet
ry processing to place the surfaces while accurately satisfying topologica
l and geometric constraints\, thus enabling subsequent parcellation and th
ickness estimation with existing methods. We present results on 5mm axial
FLAIR scans from ADNI and on a highly heterogeneous clinical dataset with
5\,000 scans.\nBiography\nKarthik Gopinath is a Postdoctoral Research Fell
ow at Athinoula A. Martinos Center for Biomedical Imaging\, Massachusetts
General Hospital\, and Harvard Medical School. His work focuses on Brain s
urface analysis of clinically acquired MR images. Previously he completed
his Ph.D. from ETS Montreal\, where his work focused on Geometric learning
for brain surface analysis\, for which he received the Governor-general a
cademic gold medal and Best Thesis award. He completed his MS by Research
from IIIT-Hyderabad. You can find more information about his work on his w
ebpage: https://sites.google.com/site/karthikharitz/\n\nHost Faculty: Dr.
Vaanathi Sundaresan\n=====================================================
=============================\nALL ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:10@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230824T123000
DTEND;TZID=Asia/Kolkata:20230824T133000
DTSTAMP:20231102T050810Z
URL:https://cds.iisc.ac.in/events/seminar-cds-102-24th-august-qameleon-mul
tilingual-qa-with-only-5-examples/
SUMMARY:{Seminar} @ CDS: #102 : 24th August: “QAmeleon: Multilingual QA w
ith Only 5 Examples”
DESCRIPTION:Department of Computational and Data Sciences\nDepartment Semin
ar\n\nSPEAKER : Priyanka Agrawal (Google Deepmind)\n\nTITLE : “QAmeleon:
Multilingual QA with Only 5 Examples“\n\nDate &\; Time : August 24\,
2023\, 12:30 PM\n\nVenue : # 102\, CDS Seminar Hall\n\n==================
=========================================================================\
nAbstract\nThe availability of large\, high-quality datasets has been one
of the main drivers of recent progress in question answering (QA). Such an
notated datasets however are difficult and costly to collect\, and rarely
exist in languages other than English\, rendering QA technology inaccessib
le to underrepresented languages. An alternative to building large monolin
gual training datasets is to leverage pre-trained language models (PLMs) u
nder a few-shot learning setting. Our approach\, QAmeleon\, uses a PLM to
automatically generate multilingual data upon which QA models are trained\
, thus avoiding costly annotation. Prompt tuning the PLM for data synthesi
s with only five examples per language delivers accuracy superior to trans
lation-based baselines\, bridges nearly 60% of the gap between an English-
only baseline and a fully supervised upper bound trained on almost 50\,000
hand labeled examples\, and always leads to substantial improvements comp
ared to fine-tuning a QA model directly on labeled examples in low resourc
e settings. Experiments on the TyDiQA-GoldP and MLQA benchmarks show that
few-shot prompt tuning for data synthesis scales across languages and is a
viable alternative to large-scale annotation.\nBiography\nPriyanka Agrawa
l is a Research Scientist at the Google Deepmind in London\, formally part
of Google Brain\, and is focused on building responsible Generative AI mo
dels and scaling them to underrepresented languages. Prior to that she was
a Senior Researcher and Lead at Booking.com and IBM Research Labs\, where
she was driving work in cross-domain transfer and representation learning
. She is also an alumni from IISc from the CSA department. Her work is pub
lished at top tier ML and NLP conferences like NeurIPS\, ACL and she holds
25+ US Patents. Priyanka also serves as Area Chair and PC member at these
conferences and has been an invited panelist and speaker at various ML/NL
P and diversity forums.\n\nHost Faculty: Dr. Danish Pruthi\n==============
====================================================================\n\nAL
L ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:9@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230824T110000
DTEND;TZID=Asia/Kolkata:20230824T120000
DTSTAMP:20231102T050821Z
URL:https://cds.iisc.ac.in/events/seminar-cds-102-24th-august-an-introduct
ion-to-artificial-intelligence-in-pharmaceutical-industry-beyond-drug-disc
overy/
SUMMARY:{Seminar} @ CDS: #102 : 24th August: “An introduction to Artifici
al Intelligence in pharmaceutical industry beyond drug discovery”
DESCRIPTION:Department of Computational and Data Sciences\nDepartment Semin
ar\n\nSPEAKER : Dr Arijit Patra is a Senior Principal Scientist at UCB Bio
pharma UK\n\nTITLE : “An introduction to Artificial Intelligence in phar
maceutical industry beyond drug discovery”\n\nDate &\; Time : August
24\, 2023\, 11:00 AM\n\nVenue : # 102\, CDS Seminar Hall\n\n==============
==========================================================================
===\nAbstract\nThe development of machine learning algorithms geared towar
ds drug discovery and chemoinformatics has seen unprecedented progress in
recent years. Much of this revolution has derived from progress in deep le
arning and generative modelling. It is worth noting\, however\, that the
‘Ideas to Patient’ journey in a pharmaceutical setting is much more ex
pansive and data-rich even beyond the discovery phases. There is a possibi
lity of deriving significant value using AI algorithms in the phases of dr
ug safety and efficacy evaluations\, patient analytics and post-market sur
veillance as these stages involve the analysis of large amounts of multimo
dal information. In this talk\, we shall discuss the potential for using m
achine learning approaches in these parts of the drug development pipeline
\, with a particular focus on machine learning for toxicologic pathology.\
nBiography\nDr Arijit Patra is a Senior Principal Scientist at UCB Biophar
ma UK. He holds a PhD in machine learning for healthcare imaging from the
University of Oxford\, where he was a Rhodes Scholar (India &\; Exeter\
, 2016). Prior to that\, he completed a dual degree in Mechanical Engineer
ing from the Indian Institute of Technology (IIT) at Kharagpur\, India. He
has also been associated with AstraZeneca\, Shell\, Microsoft Research an
d CSIR-South Africa at various points in his career and has been actively
involved in the AI4SG (AI for Social Good) community. He has authored seve
ral publications around machine learning and medical imaging and is a revi
ewer for multiple peer reviewed venues such as NeurIPS\, ICML\, MICCAI and
several journals.\n\nHost Faculty: Dr. Vaanathi Sundaresan\n=============
=====================================================================\nALL
ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:8@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230822T090000
DTEND;TZID=Asia/Kolkata:20230822T100000
DTSTAMP:20231102T050523Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-defense-cds-22-august-2023%e
2%80%b3abstractions-and-optimizations-for-data-driven-applications-across-
edge-and-cloud/
SUMMARY:Ph.D: Thesis Defense: CDS: 22\, August 2023″Abstractions and Opti
mizations for Data-driven Applications Across Edge and Cloud.”
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nPh.D. Thesis Def
ense\n\n__________________________________________________________________
________________________\n\nSpeaker : Mr. Aakash Khochare\n\nS.R. Number :
06-18-02-13-12-16-1-14050\n\nTitle : “Abstractions and Optimizations fo
r Data-driven Applications Across Edge and Cloud.”\n\nResearch Superviso
r: Prof. Yogesh Simmhan\n\nDate &\; Time : August 22\, 2023 (Tuesday)\,
09:00 AM\n\nVenue : Room No. 102 (CDS Seminar Hall)\n____________________
______________________________________________________________________\nAb
stract\nModern data driven applications have a novel set of requirements.
Advances in deep neural networks (DNN) and computer vision (CV) algorithms
have made it feasible to extract meaningful insights from large-scale dep
loyments of urban cameras and drone video feeds. These data driven applica
tions\, usually composed as workflows\, tend to have high bandwidth and lo
w latency requirements in order to extract timely results from large data
sources. Other applications may necessitate the use of multiple geographic
ally distributed resources. Such requirements may be driven by data privac
y regulations such as the General Data Protection Regulation (GDPR) of the
European Union\, need for specialized hardware\, or as a means of avoidin
g vendor lock-ins.\n\nTo support these modern applications\, a diverse com
puting landscape has emerged over the last decade. We have witnessed incre
asingly powerful Edge computing resources be available in network proximit
y to the data sources for these applications. The number of Cloud Service
Providers (CSPs) has increased along with the regions in which they operat
e. And finally\, the CSPs have supplemented Infrastructure as a Service (I
aaS) offerings with modern serverless compute offerings which promise cost
benefits as well as lower operational overheads.\n\nThe availability of c
hoices in compute resources makes it challenging for application developer
s to manage the lifecycle of their applications — from programming the a
pplication\, to optimizing it for performance\, and finally deploying it.
Typically\, developers rely on platforms that promise ease of programmabil
ity coupled with scalability with minimal developer effort. However\, the
combination of application requirements and compute resource characteristi
cs makes it challenging for platform designers to make design choices that
optimizes the application for programmability and performance. A thorough
revisit of existing platforms\, abstractions\, and optimizations is essen
tial for addressing these challenges.\n\nIn this thesis\, we tackle these
challenges with three distinct but related research contributions on scala
ble platforms\, distributed algorithms and system optimizations: (1) We pr
opose Anveshak\, a platform that provides a domain specific programming mo
del and a distributed runtime for efficiently tracking entities in a multi
-camera network\; (2) We design algorithms and heuristics to solve MSP\, w
hich co-schedules the flight routes of a drone fleet to visit and record v
ideo at waypoints\, and perform subsequent on-board Edge analytics\; and (
3) We develop XFaaS\, a platform that allows “zero touch” deployment o
f functions and workflows across multiple clouds and Edges by automaticall
y generating code wrappers\, Cloud queues\, and coordinating with the nati
ve FaaS engine of a CSP.\n\nThese platforms\, abstractions and optimizatio
ns solve different combinations of the problem dimensions\, are motivated
through real-world applications\, and the solutions are validated through
detailed experiments on distributed systems. Taken together\, this suite o
f contributions addresses the key gaps highlighted in this dissertation an
d help bridge the gap between modern computing resource characteristics an
d modern application requirements.\n\n====================================
============================\n\nALL ARE WELCOME
CATEGORIES:Events,Thesis Defense
END:VEVENT
BEGIN:VEVENT
UID:7@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230821T100000
DTEND;TZID=Asia/Kolkata:20230821T110000
DTSTAMP:20231102T045936Z
URL:https://cds.iisc.ac.in/events/m-tech-research-thesis-colloquium-cds-a-
co-kurtosis-tensor-based-featurization-for-scalable-combustion-simulations
/
SUMMARY:M.Tech Research Thesis {Colloquium}: CDS : “A co-kurtosis tensor
based featurization for scalable combustion simulations.”
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nM.Tech Research
Thesis Colloquium\n\n_____________________________________________________
_____________________________________\n\nSpeaker: Mr. Dibya Jyoti Nayak\n\
nS.R. Number: 06-18-01-10-22-21-1-19747\n\nTitle: “A co-kurtosis tensor
based featurization for scalable combustion simulations.“\n\nResearch Su
pervisor: Dr. Konduri Aditya\n\nDate &\; Time: August 21\, 2023 (Monday
) at 10:00 AM\n\nVenue: # 102 CDS Seminar Hall\n__________________________
________________________________________________________________\nAbstract
\nIdentifying low-dimensional representations of the thermo-chemical state
space for turbulent reacting flow systems is vitally important\, primaril
y to significantly reduce the computational cost of device-scale combustio
n simulations. Moreover\, these simulations are often performed to gain fu
ndamental insights into the inception of extreme/anomalous events such as
flashbacks\, flame extinction\, blow-offs\, thermoacoustic instabilities\,
etc.\, which can have detrimental effects on combustion efficiency and en
gine performance. With the scale of scientific investigations ever increas
ing\, the need for robust anomaly detection methods becomes increasingly c
ritical for judicious steering of these simulations and also aiding smooth
operations of practical engines. Recent studies have shown that the fourt
h-order joint statistical moment tensor\, i.e.\, co-kurtosis\, effectively
captures anomalies/outliers in scientific data. Accordingly\, the primary
objective of this work centers around leveraging the unique properties of
the co-kurtosis tensor to drive low-cost and scalable combustion simulati
ons and build robust algorithms for extreme event detection. Particularly\
, the first part of this work develops tools for dimensionality reduction
for chemistry\, while the second part focuses on employing a co-kurtosis b
ased detection algorithm for capturing extreme events such as flame instab
ilities in hydrogen-fired reheat burners relevant to sequential gas turbin
e engines.\n\nTo obtain low-dimensional manifolds (LDMs) that describe the
original thermo-chemical state\, principal component analysis (PCA) and i
ts variants are widely employed. An alternative dimensionality reduction t
echnique that focuses on higher order statistics\, co-kurtosis PCA (CoK-PC
A)\, has been shown to provide an optimal LDM for effectively capturing th
e stiff chemical dynamics associated with spatiotemporally localized react
ion zones. While its effectiveness has only been demonstrated based on a p
riori analyses with linear reconstruction\, in this work\, we employ nonli
near techniques to reconstruct the full thermo-chemical state and evaluate
the efficacy of CoK-PCA compared to PCA. Specifically\, we combine a CoK-
PCA-/PCA-based dimensionality reduction (encoding) with an artificial neur
al network (ANN) based reconstruction (decoding) and examine\, a priori\,
the reconstruction errors of the thermo-chemical state. We employ three co
mbustion test cases representing varying degrees of complexity in the geom
etrical domain\, combustion regimes\, ignition kinetics\, etc.\, to assess
CoK-PCA/PCA coupled with ANN-based reconstruction. Results from the analy
ses demonstrate the robustness of the CoK-PCA based LDM with ANN reconstru
ction in accurately capturing the data\, specifically from the reaction zo
nes.\n\nHydrogen’s highly reactive and diffusive nature towards decarbon
ization is prone to flashbacks\, flame instabilities\, and thermoacoustic
instabilities. For example\, in the case of reheat burners of hydrogen-fir
ed sequential gas turbine engines\, intermittent temperature and pressure
fluctuations result in flame instabilities\, such as intermittent autoigni
tion events at off-design locations that can adversely impact the engine
’s performance. To address this issue\, we develop an unsupervised learn
ing methodology based on the co-kurtosis tensor to detect the early onset
of spontaneous ignition kernels in lean premixed hydrogen combustion at vi
tiated conditions. The accuracy of the model is evaluated for various igni
tion test cases.\n\n======================================================
=========================\n\nALL ARE WELCOME
CATEGORIES:Events,Ph.D. Thesis Colloquium
END:VEVENT
BEGIN:VEVENT
UID:6@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230818T120000
DTEND;TZID=Asia/Kolkata:20230818T130000
DTSTAMP:20231102T050841Z
URL:https://cds.iisc.ac.in/events/cloud-seminar-a-carbon-first-approach-fo
r-decarbonizing-computing-prashant-shenoy-u-mass-amherst-fri-18-aug-12pm-c
ds-102/
SUMMARY:[CLOUD SEMINAR] A Carbon First Approach for Decarbonizing Computing
\, Prashant Shenoy\, U Mass\, Amherst\, Fri 18 Aug\, 12PM\, CDS 102
DESCRIPTION:========================================\nCLOUD COMPUTING SEMIN
AR SERIES\n========================================\n\nSPEAKER : Prof. Pra
shant Shenoy\, University of Massachusetts Amherst\, USA\n\nTITLE : A Carb
on First Approach for Decarbonizing Computing\n\nDate &\; Time : Friday
Aug 18\, 2023\, 12PM\n\nVenue : CDS Seminar Hall (102)\n\n===============
==========================================================================
==\nAbstract\nThe exponential growth of cloud computing has been a definin
g trend of our time\, fueled by rapidly growing demands from data-intensiv
e and machine-learning workloads. Despite the end of Dennard scaling\, the
cloud’s energy demand grew more slowly than expected over the past deca
de due to the aggressive implementation of energy-efficiency optimizations
. Unfortunately\, there are few significant remaining optimization opportu
nities using traditional methods\, and moving forward\, the cloud’s cont
inued exponential growth will translate into rising energy demand\, which\
, if left unchecked\, will translate to increasing carbon emissions.\n\nIn
this talk\, I will argue for a Carbon First approach to designing cloud c
omputing platforms and systems by making carbon efficiency a first-class d
esign metric\, similar to traditional metrics of performance and reliabili
ty. I will explain how today’s systems can be made first carbon-aware by
exposing energy and carbon usage information to software platforms and th
en made carbon-efficient by providing control over the system’s carbon u
sage. I will present several systems mechanisms to enable such carbon awar
eness and management and present several application case studies on how m
odern cloud applications can employ these mechanisms to reduce their carbo
n footprint. I will end with some open research challenges in this emergin
g field of sustainable computing.\nBiography\nPrashant Shenoy is currently
a Distinguished Professor and Associate Dean in the College of Informatio
n and Computer Sciences at the University of Massachusetts Amherst. He rec
eived the B.Tech degree in Computer Science and Engineering from the India
n Institute of Technology\, Bombay and the M.S and Ph.D degrees in Compute
r Science from the University of Texas\, Austin. His research interests li
e in distributed systems and networking\, with a recent emphasis on cloud
and green computing. He has been the recipient of several best paper award
s at leading conferences\, including a Sigmetrics Test of Time Award. He s
erves on editorial boards of the several journals and has served as the pr
ogram chair of over a dozen ACM and IEEE conferences. He is a fellow of th
e ACM\, the IEEE\, and the AAAS.\n\nHost Faculty: Yogesh Simmhan\, CDS\n\n
—————————————————————–\n\nAbou
t: The IBM-IISc Hybrid Cloud Lab (IIHCL) hosted at IISc is curating the Cl
oud Computing Seminar series with guest speakers from Industry and Academi
a speaking about the latest technologies and research on Cloud and edge co
mputing\, distributed computing systems\, and AI/ML/Big Data platforms. Mo
re details at: http://iihcl.iisc.ac.in\n\n================================
==================================================\nALL ARE WELCOME
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:5@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230811T110000
DTEND;TZID=Asia/Kolkata:20230811T120000
DTSTAMP:20231102T045245Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-colloquium-cds-modeling-phys
iological-transport-at-scales-connecting-cells-to-organs/
SUMMARY:Ph.D. Thesis {Colloquium}: CDS : “Modeling physiological transpor
t at scales: connecting cells to organs.”
DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nPh.D. Thesis Col
loquium\n\n_______________________________________________________________
___________________________\n\nSpeaker : Ms. Deepa Maheshvare\n\nS.R. Numb
er : 06-18-01-10-12-16-1-14025\n\nTitle : “Modeling physiological transp
ort at scales: connecting cells to organs.”\n\nResearch Supervisor: Prof
. Debnath Pal\n\nDate &\; Time : August 11\, 2023 (Friday) at 11:00 AM\
n\nVenue : Room No. 102 (CDS Seminar Hall)\n\n____________________________
______________________________________________________________\nAbstract\n
The physiological system is a complex network in which each organ forms a
subsystem\, and different subsystems interact to maintain the body’s ove
rall homeostasis. Within each subsystem\, functional networks exist at dif
ferent levels of complexity. The vessels that perfuse the cells within a t
issue mediate short- and long-distance communication. The ability to simul
taneously capture local and global dynamics by hierarchically bridging com
munication networks at different scales is a key challenge in holistic phy
siology modeling.\n\nWe present a scalable hierarchical framework that all
ows us to bridge diverse scales to model biochemicals’ production\, cons
umption\, and distribution in a tissue microenvironment. We developed a di
screte modeling framework to simulate the gradient-driven advection–disp
ersion-reaction physics of multispecies transport in multiscale systems. B
iochemical drift in the capillary network\, transcapillary exchange\, and
cellular reactions were modeled by translating physical space into a metam
odel with functional units. We define graph operators on the finite connec
ted network representation of the discrete functional units embedded in th
e metamodel. The governing differential equations are modeled to study the
inter-compartment dynamics of the well-mixed nodal volumes by formulating
the transport dynamics in the vascular domain and the transcapillary exch
ange as a ‘tank-in-series’ model. This allows our framework to scale t
o large networks and provides the flexibility to fuse multiscale models by
encoding imaging data of vascular topology and omics data of cellular rea
ctions to enhance systems-level understanding. Our framework is suitable f
or reducing the computational cost of spatially discretizing large tissue
volumes and for probing the effect of flow topology on biochemical transpo
rt to study structure-function relationships in tissues.\n\nNext\, we deve
loped a comprehensive and standardized data-driven modeling workflow to bu
ild kinetic models of cellular metabolism. There are significant challenge
s in developing computational models at the cellular level due to the subs
tantial variation in data across different experimental systems\, animal s
pecies\, and cell lines. We have created open\, free\, and FAIR (findable\
, accessible\, interoperable\, and reusable) assets that can be used to st
udy pancreatic physiology and glucose-stimulated insulin secretion (GSIS).
The data curation\, integration\, normalization and data fitting workflow
\, and a large database of metabolic data of the pancreatic beta-cell base
d on experimental and clinical data from 39 studies spanning 50 years of p
ancreatic\, islet\, and beta-cell research in humans\, rats\, mice\, and c
ell lines were used to construct a novel data-driven kinetic SBML (Systems
Biology Markup Language) model of GSIS in the pancreatic beta-cell. The m
odel consists of detailed glycolysis and phenomenological equations for bi
phasic insulin secretion coupled to cellular energy state\, ATP dynamics\,
and (ATP/ADP ratio). The predicted glucose-dependent response of glycolyt
ic intermediates and biphasic insulin secretion are in good agreement with
experimental measurements\, and our model predicts that the factors affec
ting ATP consumption\, ATP formation\, hexokinase\, phosphofructokinase\,
and ATP/ADP-dependent insulin secretion have a major effect on GSIS.\n\nFi
nally\, we present KiPhyNet\, an online network simulation tool connecting
cellular kinetics and physiological transport. The tool allows users to s
imulate and interactively visualize pressure\, velocity\, and concentratio
n fields for applications such as flow distribution\, glucose transport\,
and glucose-lactate exchange in microvascular networks.\n\nWhen extended f
or translational purposes in clinical settings\, the framework and pipelin
e developed in this work can advance the simulation of whole-body models a
nd are expected to have major applications in personalized medicine and dr
ug discovery.\n\n=========================================================
======================\n\nALL ARE WELCOME
CATEGORIES:Events,Ph.D. Thesis Colloquium
END:VEVENT
BEGIN:VEVENT
UID:1@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata;VALUE=DATE:20230807
DTEND;TZID=Asia/Kolkata;VALUE=DATE:20230809
DTSTAMP:20231007T192521Z
URL:https://cds.iisc.ac.in/events/cds-research-expo-2023/
SUMMARY:CDS Research Expo 2023
DESCRIPTION:We are pleased to inform you that the Department of Computation
al and Data Sciences\, Indian Institute of Science Bangalore is organizing
the CDS Research Expo on 7th and 8th August 2023. The two-day expo is jam
-packed with engaging sessions\, including plenary talks and presentations
by senior Ph.D. students\, poster sessions\, and industry-connect events.
The CDS Expo brings together researchers and faculty members of the CDS d
epartment\, where this expo serves as the best platform to research studen
ts to present their work to a larger audience\, share their ideas\, and ob
tain constructive feedback.\n\nWe cordially invite you to join us at this
event for interaction among students\, faculty members\, and top-notch res
earchers from academia and industry. Your presence would greatly contribut
e to the success of this event.\n\n\n\n\n\n\n\n\n\n\nPoster Presentation S
chedule\n\n\n\n\n\n\nOrganising Committee Members\nVaanathi Sundaresan\nRa
tikanta Behera\nChirag Jain\n\nVolunteers\nRaji Susan Mathew\nAjeya B S\nM
ohit Singhal\nTarun Nayak\nRamanujam Narayanan\nArchakam S Anudeep
CATEGORIES:Events,Talks
END:VEVENT
BEGIN:VEVENT
UID:4@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230731T113000
DTEND;TZID=Asia/Kolkata:20230731T123000
DTSTAMP:20231007T192554Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-colloquium-cds-sparsificatio
n-of-reaction-diffusion-dynamical-systems-in-complex-networks/
SUMMARY:Ph.D. Thesis {Colloquium}: CDS : “Sparsification of Reaction-Diff
usion Dynamical Systems in Complex Networks.”
DESCRIPTION:Speaker : Mr. Abhishek Ajayakumar\n\nS.R. Number : 06-18-01-10-
12-18-1-16176\n\nTitle : “Sparsification of Reaction-Diffusion Dynamical
Systems in Complex Networks.”\n\nResearch Supervisor: Prof. Soumyendu R
aha\n\nDate &\; Time : July 31\, 2023 (Monday) at 11:30 AM\n\nVenue : T
he Thesis Colloquium will be held on HYBRID Mode # 102 CDS Seminar Hall /M
ICROSOFT TEAMS.\n\nPlease click on the following link to join the Thesis C
olloquium:\n\nMS Teams link:\n\nhttps://teams.microsoft.com/l/meetup-join/
19%3ameeting_MDFmNDdlYWYtOTI0ZC00Njg5LWEwZjUtNzZhZjNiY2I0MDhm%40thread.v2/
0?context=%7b%22Tid%22%3a%226f15cd97-f6a7-41e3-b2c5-ad4193976476%22%2c%22O
id%22%3a%22f67fa1e8-f789-4048-bb53-c2880088cc5d%22%7d\n\n_________________
_________________________________________________________________________\
nAbstract\nGraph sparsification is an area of interest in computer science
and applied mathematics. Sparsification of a graph\, in general\, aims to
reduce the number of edges in the network while preserving specific prope
rties of the graph\, like cuts and subgraph counts. Modern deep learning f
rameworks\, which utilize recurrent neural network decoders and convolutio
nal neural networks\, are characterized by a significant number of paramet
ers. Pruning redundant edges in such networks and rescaling the weights ca
n be useful. Computing the sparsest cuts of a graph is known to be NP-hard
\, and sparsification routines exist for generating linear sized sparsifie
rs in almost quadratic running time. The complexity of this task varies\,
and it is closely linked to the level of sparsity we desire to achieve. In
our study\, we extend the concept of sparsification to the realm of react
ion-diffusion complex systems. We aim to address the challenge of reducing
the number of edges in the network while preserving the underlying flow d
ynamics.\n\nSparsification of such complex networks is approached as an in
verse problem guided by data representing flows in the network\, where we
adopt a relaxed approach considering only a subset of trajectories. We map
the network sparsification problem to a data assimilation problem on a Re
duced Order Model (ROM) space with constraints targeted at preserving the
eigenmodes of the Laplacian matrix under perturbations. The Laplacian matr
ix is the difference between the diagonal matrix of degrees and the graph
’s adjacency matrix. We propose approximations to the eigenvalues and ei
genvectors of the Laplacian matrix subject to perturbations for computatio
nal feasibility and include a custom function based on these approximation
s as a constraint on the data assimilation framework.\n\nIn the latter pha
se of the study\, we developed a framework to enhance POD-based model redu
ction techniques inreaction-diffusion complex systems. This framework inco
rporates techniques from stochastic filtering theory and pattern recogniti
on.\n\nGetting optimal state estimates from a noisy model and noisy measur
ements forms the core of the filtering problem. By integrating the particl
e filtering technique\, we generate the reaction-diffusion state vector at
various time steps\, utilizing noisy measurements obtained from ROM. To e
nsure the framework’s effectiveness\, we make intermittent updates to th
e system variables during the particle filtering step\, employing the care
fully crafted sparse graph. The framework is utilized for experimentation\
, and results are presented on random graphs\, considering the diffusion e
quation on the graph and the chemical Brusselator model as the dynamical s
ystems embedded in the graph. We discuss the method’s limitations\, and
the proposed framework is evaluated by comparing its performance against t
he Neural ODE-based approach\, which serves as a compelling reference due
to its demonstrated robustness in specific applications.\n\n==============
=================================================================\n\nALL A
RE WELCOME
CATEGORIES:Ph.D. Thesis Colloquium
END:VEVENT
BEGIN:VEVENT
UID:3@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230705T110000
DTEND;TZID=Asia/Kolkata:20230705T120000
DTSTAMP:20231007T192549Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-defense-hybrid-cds-05-july-2
023%e2%80%b3constrained-stochastic-differential-equations-on-smooth-manifo
lds/
SUMMARY:Ph.D: Thesis Defense: HYBRID: CDS: ″Constrained Stochastic Differ
ential Equations on Smooth Manifolds.”
DESCRIPTION:\n\n\n\nSpeaker : Mr. Sumit Suthar\n\
n\nS.R. Number : 06-18-01-10-12-17-1-14862\n\n\n\n\nTitle
: “Constrained Stochastic Differential E
quations on Smooth Manifolds.“\n\nResearch Supervisor: Prof. Soumyendu R
aha\n\n\n\n\n\nDate &\; Time : July 05\, 2023 (Wednesday)\
, 11:00 AM\n\n\n\n\nVenue : The Thesis D
éfense will be held on HYBRID Mode # 102 CDS Seminar Hall /MICROSO
FT TEAMS\n\nPlease click on the following link to join the Thesis Defens
e:\n\nMS Teams link\n\nhttps://teams.microsoft.com/l/meetup-join/19%3ameet
ing_YjhkNjhjMzktMjQ2OC00YzYwLTkxZWQtNDJiMWViZDUzYTc0%40thread.v2/0?context
=%7b%22Tid%22%3a%226f15cd97-f6a7-41e3-b2c5-ad4193976476%22%2c%22Oid%22%3a%
22f67fa1e8-f789-4048-bb53-c2880088cc5d%22%7d\n\n\n\n______________________
____________________________________________________________________\nAbst
ract\n\n\n\n\n\n\n\n\n\n\n\nDynamical systems with uncertain fluctuations
are usually modelled using Stochastic Differential Equations (SDEs). Due
to operation and performance related conditions\, these equations may als
o need to satisfy the constraint equations. Often the constraint equations
are “algebraic”. Such constraint equations along with the given SDE
form a system of Stochastic Differential-Algebraic Equations (SDAEs).\nT
he main objective of this thesis is to consider these equations on smooth
manifolds. However\, we first consider SDAEs on Euclidean spaces to unde
rstand these equations locally. A sufficient condition for the existence
and uniqueness of the solution is obtained for SDAEs on Euclidean spaces.
We also give necessary condition for the existence of the solution. Base
d on the necessary condition\, there exists a class of SDAEs for which th
ere is no solution. Since all SDAEs are not solvable\, we present methods
and algorithms to find approximate solution of the given SDAE. \nIn o
rder to extend this work to smooth manifolds\, we consider second order s
tochastic differential geometry to construct Schwartz morphism to represe
nt SDEs with drift that are driven by p-dimensional Wiener process. We sh
ow that it is possible to construct such Schwartz morphisms using what we
call as diffusion generators. We demonstrate that diffusion generator ca
n be constructed using flow of second order differential equations\, in p
articular using regular Lagrangians. The results obtained for SDAEs on Eu
clidean spaces are extended to SDAEs on smooth manifolds using the framew
ork of diffusion generators. We show that the results obtained for SDAEs
on Euclidean spaces translate to the manifold setting with minimal modifi
cations. We have derived Ito-Wentzell’s formula on manifolds in the fra
mework of diffusion generators to obtain approximate bounded solution wit
h unit probability. Another type of approximate solution is bounded soluti
on such that the probability of explosion is bounded by α <\; 1. We pr
esent algorithms to compute approximate solutions of both type. This has
been demonstrated with an example of SDAE on a sphere.\n=================
==============================================================\n\n\n\n\nAL
L ARE WELCOME\n\n\n\n\n\n
CATEGORIES:Thesis Defense
END:VEVENT
BEGIN:VEVENT
UID:2@cds.iisc.ac.in
DTSTART;TZID=Asia/Kolkata:20230703T110000
DTEND;TZID=Asia/Kolkata:20230703T120000
DTSTAMP:20231102T050419Z
URL:https://cds.iisc.ac.in/events/ph-d-thesis-colloquium-cds-03-july-2023-
end-to-end-resiliency-analysis-framework-for-cloud-storage-services/
SUMMARY:Ph.D. Thesis {Colloquium}: “End-to-end Resiliency Analysis Framew
ork for Cloud Storage Services.”
DESCRIPTION:\n\n\n\nSpeaker : Ms. Archita Ghosh
\n\n\nS.R. Number : 06-18-02-10-12-17-1-14492\n\n\n\n\nTitle
: “End-to-end Resiliency Analysis Fr
amework for Cloud Storage Services.“\n\nResearch Supervisor: Dr. J. Laks
hmi\n\n\n\n\n\nDate &\; Time : July 03\, 2023 (Monday) at
11:00 AM\n\n\n\n\nVenue : Room No. 102 (CD
S Seminar Hall)\n\n\n\n\n_________________________________________________
________________________________\n\n\n\n\nAbstract\n\n\n\n\n\n\n\n\n\n\n\n
\n\nCloud storage service brought the idea of a global scale storage syste
m available on-demand and accessible from anywhere. Despite the benefits\,
resiliency remains one of the key issues that hinder the wide adaptation
of storage services. The data is hosted on cloud data centers containing h
undreds of thousands of commodity-grade hardware with layers of complex so
ftware. Failures due to system crashes\, natural disasters\, cyber-attacks
\, etc.\, are common and frequent in such environments. To keep the servic
e unaffected by such events\, resiliency is essential for cloud systems. F
or storage services\, resiliency is far more critical because losing acces
s to data or\, more importantly\, a complete data loss can have a catastro
phic impact on the client.\n\nThe existing works on storage resiliency foc
us on maintaining sufficient user data redundancy in the system to maintai
n a reliable service. However\, providing a global-scale storage solution
requires various functional and management layers to ensure the service is
accessible and all the stored items are durable. The first part of our wo
rk proves that resiliency at the stored data level does not guarantee serv
ice level reliability. A generic cloud storage system model is designed to
analytically show that the reliability achieved at the service level dras
tically differs from the reliability ensured by stored data redundancy. Th
is motivates us to bring the entire system into purview to understand clou
d storage resiliency.\n\nDue to the complexity and variation of large-scal
e storage architectures\, assessing end-to-end storage resiliency is a cha
llenging task. To achieve this\, the second part of the work proposes a ge
neric resiliency evaluation method for cloud storage services. The method
identifies the essential functional layers for storage service and the com
ponents constituting the layers. It then performs an in-depth behavior ana
lysis during all possible failures of each component. The method is used t
o assess the resiliency of two diverse and real-world cloud storage servic
es\, OpenStack Swift and CephFS. The analysis identifies various resilienc
y weak points in the service architectures and depicts the effectiveness o
f different resiliency methods used at various layers.\n\nThe third part o
f the work extends the resiliency evaluation method to understand the corr
elation of resiliency with the service usage pattern. A storage service ca
n be used for different use cases resulting in the variation of request in
terarrival time\, read and write ratio\, accessed data and metadata\, etc.
Hence\, the components involved in access sequences may differ\, and so c
an their failure impact. Using the improved resiliency evaluation method a
nd access patterns identified from real traces\, we show that resiliency c
an be selective and dynamically adjusted based on workloads without affect
ing service reliability.\n\nFinally\, The work defines an end-to-end resil
iency analysis framework for cloud storage services that enables quantific
ation\, comparison\, and optimization of cloud storage resiliency. The fra
mework allows effective modeling of cloud storage resilience by combining
the resiliency of each component participating in service reliability main
tenance for specific workloads. The framework successfully models the resi
liency of OpenStack Swift and CephFS as Stochastic Petri Nets (SPNs). The
models are used to quantify and compare the resiliency of the above two se
rvice architectures and demonstrate how to optimize resiliency while achie
ving expected service reliability.\n\n====================================
===========================================\n\n\n\n\nALL ARE WELCOME\n\n\n
\n\n\n
CATEGORIES:Events,Ph.D. Thesis Colloquium
END:VEVENT
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
X-LIC-LOCATION:Asia/Kolkata
BEGIN:STANDARD
DTSTART:20220703T110000
TZOFFSETFROM:+0530
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
END:VCALENDAR