ICS Video Player

Faculty Mini-Talks - Gopal Pandurangan's Computer Science Faculty Mini Talks

Transcript ×

Auto highlight

Off

Font-size

00:01	Hello everyone. Welcome to this minute . I'm Gopal pondering gonna prosper in

00:06	computer science department and I lead the and distribute computing group in the

00:10	So the current members of my group PhD students Carly Durrani and trees wanna

00:16	. And we are going to soon a new post doc William moses who's

00:21	to join us soon recently we graduated PhD students. Samantha Chatterjee who is

00:26	a postdoc at texas A and m fatty with trimble and Dean Farm was

00:31	google. So the research focus of group is distribute algorithms and computing.

00:38	this has been motivated to a large in fact this whole research focus just

00:43	out around the world by many Has been motivated by this major shift

00:49	paradigm from the traditional single process computing multi process computing which is characterized by

00:56	agents and computers and networks doing the and communication for example the internet is

01:04	biggest example of a distributed communication Okay, so you have the social

01:10	, you also have cloud and data computing where thousands of servers or compute

01:16	or I mean communication entities are collectively the job to do massive crossing given

01:24	single computer, your laptop is not single computer. It's actually a multicore

01:30	which has multiple process inside the machine distribute and carol crossing. And the

01:37	of distributed computing is to study the of this, this whole new paradigm

01:42	has exploded for the last 2030 Okay. And do you want to

01:48	the foundations of distributed computing? Network and big data. Okay. And

01:53	goal is to design, analyze and efficient distribute algorithms and to sort of

01:59	the performance and see what is possible what is not possible. So

02:05	this welcomes is a very important field computer algorithms that focuses on multi processor

02:11	multi agent systems and networks. And field is characterized by fundamental problems like

02:17	pantry leader election consensus is the The same problem that underlies for

02:23	the the blocking technology that is sought , routing shortest parts designed in

02:31	Falter on competition. Again, this underlies the Blockchain. Today's Blockchain technology

02:36	example. So many of these problems the internet to various things that you

02:42	today from internet wireless networks to block to whatever. So and the goal

02:48	descriptions to study the foundation problems and best algorithms that will enable these kinds

02:56	new technologies also. Okay. And complexity measures of distribute albums are basically

03:02	and time. And my first search is basically looking at this in a

03:08	sort of classical way that we want design efficient algorithms that are optimal with

03:13	to both these complexity measures for various props that I for example, those

03:18	are listed in the previous light. . And you want to understand the

03:22	between the two measures. How do get the best albums possible that that

03:28	well with respect to both messages and . And this is supported by

03:34	That's the first project. So the project that I'm working on is on

03:38	networks. Okay. So many modern like open networks are highly dynamic.

03:44	. This includes also like adults, and wireless networks. So they're dynamic

03:49	not join and lead the network continuously . So it's not a static

03:53	So it's always change changing all the . And since these are distributed,

03:59	algorithms underlies the operation of these For example, look at Bitcoin,

04:03	actually a very highly dynamic appear to a network where users join in this

04:08	and the most important album that that of one of the most important algorithms

04:13	underlies Bitcoin is this consensus algorithm that the Cryptocurrency. Okay. So it's

04:20	distributed algorithm. And the goal of dynamic networks project is to study audio

04:29	, analyze scalable and robust and secure for fundamental products that come in dynamic

04:38	. Okay, so these problems are challenging in static networks but like Bitcoin

04:43	other Blockchain networks, I mean they dynamic. I mean like it was

04:47	proving like networks like twitter and done those things. Right. So,

04:52	how do you sort of solve these in in the dynamic setting.

04:56	so my third project is on big data. Okay, so graphs and

05:02	data are so important. I mean come all the time and there are

05:06	problems from shortest parts to minimum spanning to page rank, clustering and so

05:10	. Okay. And the big challenge big data is that computing on large

05:15	. Very large graphs. Okay. these are the kinds of graphs that

05:18	and other companies like google. I the And so there are billions of

05:24	large. I mean the sizes are and trillions of edges. Okay.

05:30	, so it's very difficult to even on these graphs because they're so

05:34	Okay. And it's also difficult to your standard graph that we learn in

05:39	algorithm scores to apply to these Okay. And so the goal of

05:46	third project is to look at distribute . Okay. So you instead of

05:50	one machine to compute to do competition these graphs, which is not even

05:55	to doesn't have enough memory to fit graph. So you distribute the input

06:02	many machines. And also you do complicated computation. Okay. So you

06:06	to compute something on this graph? each part of the graph is a

06:10	piece of the machine. I there's different machine and you jointly compute

06:16	on this. I mean, any on this graph. Okay. And

06:22	the goal is to design and study implement efficient district relevance for large scale

06:27	problems, for example, this can like shortest parts or counting the number

06:31	triangles and so on and so I mean there are many problems that

06:35	the graph problems that are very important many applications. And this is sort

06:41	also the underlies graph mining and so and so forth. So for

06:46	one important application is to competition epistemology this is relevant to the current covid

06:52	situation. So recently, in fact wrote a paper that that looks at

06:58	algorithms in that in the epidemiology especially with respect to covid. So

07:04	can find a paper on the So this is supported by menace of

07:07	data grant. Okay. So the summarize the main goals of my research

07:13	is to tackle important problems and disabled panel computing and networks and of course

07:19	always want to have practical applicants in and we have been pretty successful in

07:25	in top conferences and journals in the in this areas. So the relevant

07:30	that you want to take if you to sort of join and work with

07:37	is of course data section albums, is sort of the bread and

07:41	I'm teaching that now this fall distribute that I'll teach sometime hopefully in the

07:46	year or so. And also random which I'm also going to teach in

07:49	next year or so. Okay, besides these courses, I mean other

07:54	that are offered by my colleagues are quite interesting and important. So like

08:00	data analytics, panel competitions, computer and machine learning. Okay, so

08:04	you have any questions, feel free contact me, so thank

Previous Next

00 : 01