© Distribution of this video is restricted by its owner
Transcript ×
Auto highlight
Font-size
00:00 the director of the pattern analysis laboratory the idea of this laboratory is to

00:06 artificial intelligence and machine learning techniques in applications. We cover a lot of

00:14 areas but we mainly focus on applications physics and astronomy. Now what we

00:22 is always the result of uh collaborative . Um I want to acknowledge that

00:28 have very bright and hardworking students that crucial to to attain good results in

00:35 laboratory. Okay. Right. So to position our research lab in in

00:42 in a specific area of study, do artificial intelligence and sometimes we we

00:50 use of certain areas within artificial intelligence outside machine learning such as search techniques

00:57 knowledge representation. Uh but we mainly on on what is known as machine

01:04 . And with machine learning we do and regression and clustering mainly some reinforcement

01:13 and genetic algorithms now are areas of within machine learning fall mainly in what's

01:21 transfer learning. The problem of building model in certain tasks. For example

01:28 astronomy you want you may want to how to classify certain types of stars

01:34 when a new survey, astronomical survey , then you realize that your previous

01:39 is no longer applicable because of the data because of the new instrumentation and

01:45 on. And the question is how you adapt the previous model to the

01:49 task uh to leverage that knowledge and you gain before avoiding having to start

01:57 scratch. So this area of transport is also known as the main

02:03 Uh Similar area called metal learning is how to create learning systems that that

02:11 and adapt over time and learn how correct a based learning system. So

02:17 learning to learn the learning system now how to correct itself automatically.

02:25 we've also been exploring to new One is called symbolic learning and the

02:29 one is called Discovery Systems and I'll more about this. As I

02:34 we have many applications of machine We we do use deep learning and

02:41 I mentioned, transfer learning domain imitation metal learning to areas in astronomy,

02:49 science. Lately we've been working on main research collaborations with astronomers uh here

02:58 the U. S. And in . One of the projects is about

03:02 super knobby caesar is a massive explosions are crucial because they help us by

03:10 these explosions and and the light it to measure the distance to different

03:16 And also we were working on a identifying Boyd's in the large structure of

03:25 universe using machine learning techniques. I that we have new areas here of

03:32 . 11 of them is called symbolic within the whole area discovering systems,

03:39 equations. There's been a hype there uh on how to use machine learning

03:43 discover new equations because we focus in and astronomy. But this applies to

03:50 areas in in basic science like biology chemistry. And the idea is that

03:55 you have certain data um and certain about, you know, how the

04:00 was gathered, What are the laws that that that area where the data

04:07 was gathered? Data is how can use machine learning to come up with

04:13 equations that explain the data? So whole idea is how to discover new

04:18 in science. Alright? So if you're interested at any point to join

04:26 lab, just send me an I'll be happy to discuss that with

04:29 . Thank

-
+