Okay, guys this is Mail Bag week number 
two we have a bunch of things for you 
you'll see not only the Mail Bag, but you 
will see a couple of new lectures that 
are, are intended to ease some of the 
pain when you are moving from assignments 
to assignments so essentially we're 
going to update you on the, on your 
progress in aggregate terms, okay? 
And then we'll go into some of the more 
specific topics, some of the questions 
that you guys ask, and, and give you, you 
know, more hints on which video to look 
next if you have some different kinds of 
issues, okay? 
So this is the number of students active. 
We have about 11,000 students active 
right now, this week. 
You know it was 17 last week its a you 
know a normal decrease at this point not 
worry about that, this is also something 
that is interesting to report this is 
essentially the, the prolonging languages 
that people are reported to know during 
one of the survey its very interesting to 
see how popular Python has become, okay? 
And obviously a lot of you are using 
Python as the implementation language 
here. 
so this is an interesting statistics. 
it was kind of surprising to me. 
Especially the number of people who know 
MATLAB compared to other languages. 
video views 11,000 this week of the 
introductory video, okay? 
So this is pretty good. 
The, you see the numbers here? 
For CP, local search you know 
mathematical programming. 
Very healthy, it's normal. 
You know, we are still looking, we are 
still at the beginning of the class. 
We are not expecting people to look at 
these right now. 
you know, the assignments are also 
prgoressing in each of these topics. 
4,000 views of the mail back last week. 
This is good you know, a little, you 
know, people could, you know, actively 
look at this. 
because this is giving also a lot of 
hints on how to actually approach the 
classes. 
And answering kind of questions globally 
as well. 
this is the students active in the class, 
okay? 
in in assignments. 
And once again, you know, you have a 
drop, okay? 
which is roughly the number, the same 
drop as active students [INAUDIBLE]. 
Okay what you see here is interesting, 
So, the number of submission which have 
been graded has a very tiny drop, 
[INAUDIBLE], I mean, it's a drop, but 
it's not huge, okay, which is also good. 
People are, you know, basically, moving 
along to the next assignments, which is 
very nice. 
And this is the number of submission by 
active students you see that actually the 
students who are active are actually 
submitting more which is, which is also 
great, okay. 
So I'll give you some more graph in a 
moment, but this is the big picture at 
this point, okay this is also the 
participation we've assignment what you 
see is two column the, the maximum number 
of submission for a [UNKNOWN] plotting a 
problem and the minimum ones. 
you see that knapsack is about you know 
the, the least you know 2, you know 
2.5000 students submitting are submitting 
the various the, the various solutions 
the maximum is 3, 3000 which is good, 
okay, so very healthy also are the number 
for you know colouring its about close to 
a 1000 if you could raised that a little 
bit that would be terrific. 
And you see that some people have 
actually already, you know solved all the 
assignments and some of the puzzle 
challenge as well, okay? 
But once again, we don't expect that, you 
know. 
This is a week by week class, okay? 
We give you all, you know, you can do 
everything you want but we, we assume 
that people are going to progress week by 
week on every one of these assignments, 
okay? 
no for the grading, once again i want to 
set the expectation right,so, so getting 
all 10's in this class is amazingly 
difficult. 
Ok so, so we are not expecting you know? 
a large number of students to actually 
achieve that This is not their intent. 
That would be really really terrific, 
okay? 
But this is, this requires a lot of work. 
This requires a lot of knowledge. 
And even that. 
That requires a lot of time just getting 
these solutions on every one of them. 
So we don't expect you to do that, in a 
sense. 
So what we want the class to do is focus 
on the seven first. 
You know, solve the assignments. 
Get seven on everything. 
You get a. 
Certification at the end and that is 
really feasible, you know, as far as we 
can judge, okay? 
So the bar is set reasonably for getting 
7's on all of the parts.OK? 
Now what you can do is once you get these 
7's, once you have allegories, well 
several allegories for popular 
assignments, and you get 7 on them, you 
can say oh no I want to proof. 
That's good, that's healthy and this is 
what we want you to do. 
But don't focus on 1 particular thing, 
and say, oh I don't get a 10 on that one. 
Okay, so learn more techniques inside 
during the class. 
And you can always go back to these 
assignments, and correct them. 
We lea, we give you some weeks at the end 
to actually do that. 
Okay, so in a sense, guarantee you're 
doing craft coloring and let's say you 
haven't look at the local set lecture, 
you can go, you know, let's say five 
times all of the six party [UNKNOWN] but 
the last one you can't, you will see 
techniques later on, you can go back 
there and maybe get a 10 on that one. 
Okay, that is the essence of the class, 
right? 
So so you don't have to focus on getting 
every ten, every you know, a ten on every 
part of the assignment before moving to 
the next assignment, don't do that okay. 
so there are basically two approaches 
here that you can always do when you want 
to improve right, so you can have the 
scaleable solution right that I was just 
talking about, trying to get seven 
everywhere Okay, and when you get that, 
you basically get, you know, this is 
basically trying to find an approach 
which scales, well on all the parts 
essentially, on all the instances. 
If you can get that, it's not going to be 
the best algorithm, but you get a good 
grade, right. 
So, you get, you know, essentially 7, on 
the 6 part you get 42 for every one 
assignment. 
The other way to do it is to say ooh, I 
really want some of these 10's, I want to 
prove [UNKNOWN] on these problems and 
it's possible, right. 
And when you do that, you may say, oh but 
that approach doesn't scale as well as 
let's say, another approach, okay? 
But that's fine as well. 
You could say, okay, so on four of these 
[UNKNOWN] I get a ten, on the last two... 
I get a 3 because its not scaring you get 
a still 4 or 6 which is a very nice grade 
right, and then later on in the class you 
can say oh but i have a little bit more 
time ,this is the last two weeks, i could 
see if i can implement something else to 
raise that bar, so that's also something 
you can do,so these two approaches are 
what we want you guys to do and we will 
say how you can get there in some of the 
new video that i am going to you know 
mention in moment. 
Once again getting 10 on everything is 
really hard, okay? 
that's why we have a specific 
distinctions for that particular, for 
those people who could actually do that 
,so in the sense don't focus on getting 
10 in everything. 
Okay, so its just very very hard and 
graphically in particular in very very 
hard So, this is a curve which is really 
interesting. 
What you see, so this, this is 
essentially you know seven of the, the 
red line that you see there is getting a 
seven of average on all the parts that 
you are assuming to have completed by 
now, okay? 
So, what you see is that there are about 
800 students, you know 
You know, on the right of this curve so 
those people are on track. 
What you see also is a large body of 
students that are very, very close and 
our goal is to take these guys and bring 
them over the curve, okay? 
And so, those people who are there they 
just don't have to worry so look at the 
stuff that we are saying today, look at 
the stuff that we present in the 
lectures, and that should allure you to 
go over this bar, okay? 
Sending expectation right, don't get to, 
don't try to get that. 
Okay, so the goal is to have a seven or 
average. 
Okay, you don't have to be the best in 
everything. 
Okay, so you will have more time later on 
to actually refine those assignment and 
that's the goal here right. 
So, we want to bring you over this curve, 
and you will some of the techniques that, 
some of the strategies that we are 
following to actually do that in 
practice, okay? 
So, we want these people to move, you 
know, across this curve and that's the 
goal, that's the goal of the couple of 
lectures that we are going to present. 
Okay, now there are a lot of questions 
that, you know, lot of questions on the 
board and people are very good at 
responding to them.They are three of them 
that always, you know, come back Okay, 
and we want to address them. 
The first one is why is this class so 
hard, then you know, I am completely 
lost, how do I get started, and we love 
the [UNKNOWN] as well. 
Okay, and what is really the goal of this 
class, okay? 
Okay, and what is really the goal of this 
class, okay? 
And so I think, you know, we basically 
specify those things at the beginning but 
I want to go over again. 
Over them again, okay? 
So, why is this class so hard? 
It's the basic principle that we have, 
this is the way we want to teach. 
So there is this beautiful, you know, 
Beatles song, I want to hold your hand. 
That's not what we are about, we don't 
want to do that. 
We don't want to give you the algorithm 
for graph coloring, and tell you, 
implement it, okay? 
So, because you don't learn as much as 
saying this other principle, okay? 
You can apply them to all these problems, 
okay? 
But now, you have to find the one that's, 
that is right for this particular 
problem. 
And you have to actually [INAUDIBLE] it 
yourself, so that. 
you can actually understand, or you put 
this [UNKNOWN] in practice, okay? 
And so, for me, you know the best way to 
do that, is actually get your hands 
dirty. 
Don't get to many guidelines, because 
when you discover things yourself, there 
are two things that happens. 
First you are going to retain them much 
better. 
And the second thing, is it's going to 
boost your confidence. 
And you know that if you see a problem 
later on Oh, I have done it for graph 
coloring. 
I know I can do it for, let's say, time 
tabling. 
Okay? 
So, in a sense, what we want is these two 
things. 
We want you to retain doing formation 
bets better. 
And by doing it yourself, you will retain 
that better, okay? 
And then the second thing that we want, 
okay? 
Is that you are also able to, to. 
in a sense, apply this much more widely 
later on, okay? 
And to do that, you have to have the 
confidence that you know that if, you 
havn't seen that problem, I have not to 
much guidance and I can apply it. 
Now I want to tell you a story. 
So I got educated in Europe. 
Most of my studies were memorizing 
things. 
And there were 2 or 3 classes that were 
not like that, and one of them was just 
crazy, right? 
So we were graded on 20 and that class, 
all the students had above 16, okay, 
well, part of the students had about 16 
and part of the students below 4, you can 
imagine, right? 
And it was like, more like 20%, 80%, so 
it was crazy, right? 
So this is the first time the professor 
was actually giving that class, okay? 
he was not able to actually give that 
kind of grade later on, he got, you know. 
Told that, you know, this is not the way 
you should grade. 
But anyway, so with that class, what's 
really, really transformally for me, 
alright? 
So it's just the only thing that that 
class was trying to tell you is that. 
You know, use your mind, try to solve the 
problem, okay? 
So we try to do that, but a much weaker 
extent in a sense, right? 
So here we are giving you a lot of 
guidance. 
But not sufficiently so that we give you, 
you know, an [UNKNOWN] that you have to 
quote. 
So, this is the spirit of this class 
Okay. 
So, the second thing that we want to do 
is for you to experience NP-hardness 
Okay, so we are not cheating any 
[UNKNOWN] Okay. 
So, when we give you graph coloring this 
is an amazingly hard problem Okay. 
So, we did it on purpose right. 
So, graph coloring is where you say these 
problem are really, really, really 
difficult, I can't prove up optimality on 
you know, the largest instance, okay? 
And this is on purpose. 
We want to show you that this is really 
really hot, okay? 
And the only way we can show you that is 
by you experiencing it, okay? 
You try it. 
If I tell you why would you believe me, 
okay? 
And in the sense if I only give you 
instance which are turkey notes, okay, 
you can solve them optimally and then 
it's not even fact, okay? 
So, here we want you to really experience 
you know these are really hard problems, 
okay? 
And problems in practice are, you know, 
rough coloring, you know, is very hard 
but problems in practice are often that 
hard. 
Okay, and You have to be really clever in 
how you actually approach them. 
So, in a sense we want you to really 
experience this thing. 
Okay, there is something called, you 
know, NP, NP-hardness. 
And this is tough okay. 
And so that was the goal of the, of the, 
the graph coloring assignment, okay. 
And once again, as I told you, we want 
you to be able to apply these ideas to 
another problem, okay. 
And so in a sense what we do, we covered, 
you know, the paradigms but we don't 
cover the real implementation. 
That's for you to discover, that's the 
basic principal of this class. 
And then we cover modeling techniques, 
but you will have to creative in choosing 
the one you apply for popular [UNKNOWN]. 
So that's the basic idea. 
And that's why its so hard, you have to 
discover a lot of things yourself. 
We don't take your hand, okay, now, so we 
understand that, you know, you know 
sometimes it's really, it's really, you 
know, it's really hard for different, for 
different people in the class or we 
understand that the diversity here is 
very large compared to some of the 
classrooms in which we have been teaching 
similar classes. 
And so, what we going to do is to, to 
actually try to help you or help some of 
you actually in, in covering a little bit 
in more detail some, some of the 
techniques that you know we basically 
assume people would know, okay and so we 
have three new lectures that I am 
going to talk about and please watch them 
if you are just you know on the, on the 
on the left of this curve. 
On the right of this curve? 
Well, whatever. 
You know, on the wrong side of that curve 
and we want to bring you along on the 
left side of the curve. 
Okay, so look at this because they're 
going to help you, okay? 
So we have two, two, two new techniques 
which are greedy, you know, one and two. 
That's basically telling you this is 
where you should start from, okay? 
So start from a greedy algorithm. 
And that will help you understand the 
nature of the problem and will give you 
some guidance on this okay, and then, you 
know what we want to talk a little bit 
about you know, if you, if you want to 
start on this assignment, this is a plan 
okay this is really you know, if you are 
you know not sure on how to approach 
them, you know, use this strategy okay 
and so that's the two you know three 
lectures I'm going to cover. 
in a moment, okay? 
Now, learning goals of the class is, is 
the class you know trying to teach you to 
use a solver or to learn how to build 
one. 
It's, it's, it's neither of them but it 
gives you insight how to, how to do this 
but the real, the real message from this 
class is that if you have an optimization 
problem, so, these are some of the most 
successful methodologies to solve them 
and we want you to understand that. 
And we want you to understand them so 
that you can actually code these 
algorithms yourself, okay? 
So you have enough information 
afterwards, you have enough informtaion 
to trnaslate that into code. 
Of course it's hard, but you have the 
information to do it, okay? 
And you can always use the forum and 
discuss with other people. 
The collaboration policy here is very 
lenient, okay? 
And then but, but that's the essence you 
could say oh I want to implement 
something in constraint programming and I 
really now how to do it, okay and now you 
implemented and then discovery oh, that 
what it takes and so on, that's one of 
the big goals of the, of the, of the 
class now it also give you the ability to 
use solvers you know in, in appropriately 
so we'll talk about that in, in some of 
the when, when we talk about the way to 
approach the assignments, okay But 
essentially a solver is only as good as 
your understanding of the various 
methodologies, okay? 
So you can put things in a solver and 
they can, the solvers can run forever if 
you're not mothering the problem right. 
Okay? 
And so in a sense, you know, we, we, you 
have to understand really how the various 
techniques are working so that you can 
use these solvers in an appropriate 
fashion.In a sense think about it here, 
you know when you look at solver we give 
the principles you can see through the 
layers and decide if that's going to be a 
good solver for your problem or not, how 
you should model your problem and so on 
Okay, so we'll come back to that because 
you know, some of the solvers you can 
take off the shelf and try to apply to 
the problems in this class won't be 
enough to get you a ten okay, so I'll 
come back to that later on okay. 
But it give you the ability to use a 
solver and then you can extend the solver 
so you can you know tailor them so that 
they work for your particular problem, 
okay. 
So, also implementing a solver is a 
completely different ball games but, once 
again you will have the principles and 
the knowledge to actually go deeper and 
look at the right paper see if you want 
to ever want implement a solver like 
this, okay? 
So, we give you the tools for doing a lot 
of different things. 
But the core here is understanding the 
methodologies at a level where you will 
you know, be able to go deeper in many 
different areas of optimization, whether 
it is application, whether it is solver, 
whether it is more, you know, 
sophisticated algorithm, okay? 
yes. 
Okay, so this is important, okay, I want 
to cover this. 
So if you use a solver, so, so some 
people may say, ooh, but the guys who 
actually know solver have a complete 
advantage on this class. 
Okay? 
So first, I don't think it's true. 
Okay? 
Because if you code, you could, you can 
do two things in this class. 
You can code the algorithm yourself. 
And you can actually get a certificate. 
It's going to take you a little bit, you 
know, longer than somebody who knows the 
solver. 
Okay? 
But you will learn a lot, okay? 
So you learn more than just, you know, 
learning a solver and then using it in a 
naive fashion, okay? 
So in a sense, yes, these people will get 
a, you know, let's say get a certificate 
in an easier fashion. 
You can do it without knowing these 
solvers, okay? 
But the really important point is that if 
you look at the way the assignments are 
chosen, okay, they have been chosen such 
that If you taken off the shelf solver 
use it in a naive fashion, okay or even 
in a sophisticated fashion is going to be 
really difficult for you to get 10 on all 
the assignments, okay. 
If you know one solver try to apply it to 
all the assignments its going to be 
tough, okay? 
Okay so in a sense what I am telling you 
here is that if you want to [UNKNOWN] 
many of these assignments You'll have to 
really creative and go beyond you know 
this, you know just a naive or a simple 
use of the solvers. 
You need to combine you use them in a 
clever fashion or you will need to use 
sophisticated technique on top of them or 
inside them and things like this, okay? 
So once again you know, yeah the reason 
advantage is viewing a solver but its 
minimum. 
Okay, and if you want to get ten on some 
of these assignments you have to go 
beyond what the solver do or use the 
solver really creatively Okay? 
So, so, so this is just to give you that 
in perspective. 
There is not that much of an advantage to 
use a solver. 
It becomes more an advantage on the 
final, on the final assignments, but even 
in those cases, okay? 
What you have to understand is that you 
will have to use this solver in a very 
clever fashion. 
It's just not knowing the solver that's 
going to help you, you have to know the 
material as well, okay? 
So, what next. 
coloring was hard. 
It was intended to do that. 
You have experience in p-hardness. 
With with with, with this, okay? 
So, it's, it's, it's, it's a very 
interesting problem because it's very 
easy to state, and it's incredibly hard, 
OK? 
And this is why this assignment is 
positioned there, OK? 
But it's, it's essentially a shock and 
it's, it was intended to be a shock, 
right? 
So the traveling salesman problem is 
going to be very different. 
Finding feasible solution is going to be 
really easy, but there are many, many 
different tools, okay? 
And here is going to be whole world can 
you optimize. 
Okay, it's the, as, as I said in the 
lectures, it's the most, you know, well 
known optimization problem, okay. 
and, once again, you know, I think you 
should focus on getting it, getting seven 
first, and then trying to see if you can 
turn them into a ten, okay. 
It's really tough to get, you know, 
optimal solution of some of these 
benchmarks as well. 
It's doable, but it's very tough, so, you 
know You know, be reasonable in the 
standards that you have and then try to 
compete afterwards when you have more 
time. 
Okay? 
So, the TSP is discussed in a lot of 
details in the local search and in the 
MIP lectures and you can use some of 
these techniques while solving it. 
Okay. 
So, keep up with good work and you know, 
it's really fun looking at, you know, 
what you guys are doing and look at the 
three lectures if you are interest, you 
know, on the left of that curve and we 
want to bring you on the right. 
Okay? 
Great, thank you.