ML Class

[Ciaran Whyte]

I don't know if people are doing the machine learning class or not but
I was thinking we could use the same group to cover both?

[Isaac Clerencia]

On Fri, Oct 14, 2011 at 12:15 AM, Ciaran Whyte <ciaran...@gmail.com> wrote:
> I don't know if people are doing the machine learning class or not but
> I was thinking we could use the same group to cover both?

Sounds reasonable. I don't think it makes sense to split it if there
is just a few of us.

--
Isaac Clerencia
isaac.c...@gmail.com

[Ciaran Whyte]

How are you getting on with the homework? 

I going to blame it on tiredness but can'y seem to implement the gradient Descent function

[Jimmy]

Yeah I'm doing that too.

I'm a bit behind because of work, playing catch up at the moment. Only
starting problem solving on the A.I and half way through Linear
Regression (one var) in ML.

Massive difference between the two, A.I seems simple and ML a bit
harder. I think I'll do the Linear Algebra bit before completing the
rest of ML, because I'm finding myself googling basic math concepts.

Is anyone having any problems with anything or is everyone cool for
now?


On Oct 14, 12:18 am, Isaac Clerencia <isaac.cleren...@gmail.com>
wrote:

> On Fri, Oct 14, 2011 at 12:15 AM, Ciaran Whyte <ciaran.wh...@gmail.com> wrote:
> > I don't know if people are doing the machine learning class or not but
> > I was thinking we could use the same group to cover both?
>
> Sounds reasonable. I don't think it makes sense to split it if there
> is just a few of us.
>
> --
> Isaac Clerencia

> isaac.cleren...@gmail.com

[Isaac Clerencia]

On Fri, Oct 14, 2011 at 12:22 AM, Ciaran Whyte <ciaran...@gmail.com> wrote:
> How are you getting on with the homework?
> I going to blame it on tiredness but can'y seem to implement the gradient
> Descent function

I did it this afternoon. I got stuck a few times but grabbing paper
and pen and drawing a few matrices helped a lot.

--
Isaac Clerencia
isaac.c...@gmail.com

[Ciaran Whyte]

I think i'm going to have to spin through some of the video's again and see what i'm missing. 

Its half 12 now, and I'm up at seven, going to call it a night. 

[Ciaran Whyte]

Hey guys i need peoples help, I came back today to start again on the
ml homework

But now none of my .m files are running in Octave

for example warmUpExercise.m

function A = warmUpExercise()
%WARMUPEXERCISE Example function in octave
% A = WARMUPEXERCISE() is an example function that returns the 5x5
identity matrix

A = [];
% ============= YOUR CODE HERE ==============
% Instructions: Return the 5x5 identity matrix
% In octave, we return values by defining which
variables
% represent the return values (at the top of the file)
% and then set them accordingly.
A = eye(5);
% ===========================================
end

when i run it in Octave i get

>> warmUpExercise()
error: `warmUpExercise' undefined near line 8 column 1
>>

But there are no errors in it? Any ideas?

On Oct 14, 12:28 am, Ciaran Whyte <ciaran.wh...@gmail.com> wrote:
> I think i'm going to have to spin through some of the video's again and see
> what i'm missing.
>
> Its half 12 now, and I'm up at seven, going to call it a night.
>

> On Fri, Oct 14, 2011 at 12:25 AM, Isaac Clerencia <isaac.cleren...@gmail.com
>
>
>
>
>
>
>
> > wrote:
> > On Fri, Oct 14, 2011 at 12:22 AM, Ciaran Whyte <ciaran.wh...@gmail.com>

> > wrote:
> > > How are you getting on with the homework?
> > > I going to blame it on tiredness but can'y seem to implement the gradient
> > > Descent function
>
> > I did it this afternoon. I got stuck a few times but grabbing paper
> > and pen and drawing a few matrices helped a lot.
>
> > --
> > Isaac Clerencia

> > isaac.cleren...@gmail.com

[Ciaran Whyte]

I've tried installing it like 3 times over from different downloads

[Isaac Clerencia]

On Sat, Oct 15, 2011 at 1:23 PM, Ciaran Whyte <ciaran...@gmail.com> wrote:
> I've tried installing it like 3 times over from different downloads

Hi there, just back from a trip to the North.

All the files that you write are run through the "ex1.m" file. You
just have to write "ex1" in the octave prompt and it should run the
different files. The ex1 file also sets up some variables that are
then used by the other files, so it's not possible to run the other
files independently unless you set up the variables manually or
something.

--
Isaac Clerencia
isaac.c...@gmail.com

[Ciaran Whyte]

I got sorted in the end, i was using Notepad++ for my editor and it was saving the files in something that was ASCII format. Once i changed the files everything just worked

Did anyone have trouble with this weeks assignment?  If so, now that the deadline has passed I think we should be allowed to share answers/explain them. What do people think?

[Brainfreeze (Lego Guru)]

The deadline for AI has been extended another day. Everyone submitted their homework at the same time (last min) and the servers crashed.  :D

The deadline for ML is gone though, and I foolishly forgot to do the Linear Algebra review questions. Because the lecturers are marked as "optional". I thought the review questions where optional also.

[Ciaran Whyte]

Right so, we'll hold off for another day

[Brainfreeze (Lego Guru)]

The answers are up for AI.

It's safe to say I bombed all search tree related questions. I mean by a large margin

How did everyone else do?

[Ciaran Whyte]

82%  - i miss counted the nodes in the search questions and thought that the loaded coin was partially observable

[Isaac Clerencia]

On Tue, Oct 18, 2011 at 10:09 AM, Ciaran Whyte <ciaran...@gmail.com> wrote:
> 82%  - i miss counted the nodes in the search questions and thought that the
> loaded coin was partially observable

96%, also failed the partially observable 'loaded coin'
--
Isaac Clerencia
isaac.c...@gmail.com

[Brainfreeze (Lego Guru)]

Where does it give you the percentage?

The loaded coin was partially observable though, you got that right if that's the case. Both sides are not observable at the same time, and it has to rely on memory to keep track of the current score to predict the next flip, as all past moves are immediately no longer observable as soon as the next one is initiated.

[Ciaran Whyte]

Click on 'Progress'

[Isaac Clerencia]

On Tue, Oct 18, 2011 at 10:25 AM, Brainfreeze (Lego Guru)
<br41n...@gmail.com> wrote:
> Where does it give you the percentage?

In the 'Progress' tab, there you can also expand each test and see the
scores for each section.

> The loaded coin was partially observable though, you got that right if
> that's the case. Both sides are not observable at the same time, and it has
> to rely on memory to keep track of the current score to predict the next
> flip, as all past moves are immediately no longer observable as soon as the
> next one is initiated.

Yeah, I answered 'not partially observable', which is wrong as you explained.
--
Isaac Clerencia
isaac.c...@gmail.com

[Brainfreeze (Lego Guru)]

Found it 42%, :-|

I genuinely thought A.I was the easiest out of the two. Leasson learned, pay more attention.

[Ciaran Whyte]

The questions were a bit up in the air tho, subject to debate and that

Plus its all about the final exam. I feel like the AI class doesn't hold up with the machine learning class at all, but it might get harder. Give us something we could go off and use

[Isaac Clerencia]

On Tue, Oct 18, 2011 at 10:33 AM, Brainfreeze (Lego Guru)
<br41n...@gmail.com> wrote:
> Found it 42%, :-|
> I genuinely thought A.I was the easiest out of the two. Leasson learned, pay
> more attention.

I think it's easier but as you say a lot of attention is required in
the quizzes. I really prefer the way the ML exercises work.

--
Isaac Clerencia
isaac.c...@gmail.com

[Tony O'Connor]

I got 72% myself.  The errors were distributed over the questions, so I think it's an issue of me not putting in enough time.  

The next set of lectures look more interesting, so I'll try to put in more effort.  ML will always be the priority for me however. 

2011/10/18 Isaac Clerencia <isaac.c...@gmail.com>

[Brainfreeze (Lego Guru)]

Another mistake I made, apart from the tree search questions, was I selected "continuous" for a few things when none of them are.

For example I thought the maze, the amount of actions would be infinite, someone could walk in a circle and never reach the destination etc. Or someone could flip the coin an infinite amount of times even the probability is one out of 10, since it had an element of randomness in it, it's not guaranteed. 

That was my logic behind it anyway, I over thought it a bit as it's wrong. So I guess a problem is only continuous is if you can see no solution to the actions ending? Rather than me making up ways to make it not end?

[Tony O'Connor]

I made the same mistake with the maze.  The way I now like to think about it is that if something is finite in the number of actions/choices you can make, it's not continuous, but discrete.  In the maze, you can only turn left, right, and go forward I think, so that's a finite number of actions or choices.  

2011/10/18 Brainfreeze (Lego Guru) <br41n...@gmail.com>

[Isaac Clerencia]

On Tue, Oct 18, 2011 at 11:27 AM, Tony O'Connor <tocon...@gmail.com> wrote:
> I made the same mistake with the maze.  The way I now like to think about it
> is that if something is finite in the number of actions/choices you can
> make, it's not continuous, but discrete.  In the maze, you can only turn
> left, right, and go forward I think, so that's a finite number of actions or
> choices.

Exactly. Without the clarification that was later posted, the problem
could have been continuous, if the robot could have chosen any angle
to move forward instead.
--
Isaac Clerencia
isaac.c...@gmail.com

[Ciaran Whyte]

Good old reddit, this may be something to move onto after the ml class is done

"Comparing with the ml-class schedule it seems to be an advanced course. It is only partially completed though."

http://openclassroom.stanford.edu/MainFolder/CoursePage.php?course=ufldl

[Ciaran Whyte]

Looking at the ai-class numbers its gone from 145k sign up to  >2000 actually watching  the vod's. Only 1000~ did the homework

I'm amazed that people are dropping out like this

[Isaac Clerencia]

On Tue, Oct 18, 2011 at 1:18 PM, Ciaran Whyte <ciaran...@gmail.com> wrote:
> Looking at the ai-class numbers its gone from 145k sign up to  >2000
> actually watching  the vod's. Only 1000~ did the homework
>
> I'm amazed that people are dropping out like this

Wow, the drop rate is indeed impressive.
--
Isaac Clerencia
isaac.c...@gmail.com

[Ciaran Whyte]

1/145  =~ 99.31% drop out rate, that must be a new record

[Ciaran Whyte]

Thats makes our study group 0.5% of the remaining class (alright i'm done with the stats now)

[Brainfreeze (Lego Guru)]

How that's incredible, now I want to make sure I finish it!

What about ML? 

[Brainfreeze (Lego Guru)]

wow*

[Ciaran Whyte]

ML doesn't use youtube so I can't tell how many people are doing it. I can see less people going to the ML class but more of those people keeping with it.

[Brainfreeze (Lego Guru)]

How do you know how many people did the homeworks?

And those views, if people watched the vid twice (I did with some), then that means even less people watched it.

[Ciaran Whyte]

http://www.reddit.com/r/aiclass/comments/lfyq3/can_we_know_how_many_people_attempted_1st/

Yeah I'd say you're right  -  I'm going to stick with it till the end

[Brainfreeze (Lego Guru)]

Is this study group private or public?

[Ciaran Whyte]

Public, as far as i know? why?

[Brainfreeze (Lego Guru)]

No reason, just wondering.

Apparently 46,000 actually did the AI homework. They confirmed. 

[Ciaran Whyte]

Awesome news, the bigger the count the more likely they'll be to do more like it in the future

[Brainfreeze (Lego Guru)]

That's what I was thinking, 

If it had a low turn out they might of considered it a failure and a waste of resources.  They days of getting high standard education online is getting closer and closer. Sure we could of all tried to learn this stuff ourselves by downloading books or reading tutorials like we normally do, put you'd end up putting it on the long finger. This sort of structure with set deadlines motivates you to keep going like physical onsite education.

[Ciaran Whyte]

How are we all doing? Haven't heard a peep out of you guys in a week,
Any one drop out yet?

On Oct 19, 3:03 pm, "Brainfreeze (Lego Guru)" <br41nfr3...@gmail.com>
wrote:

> That's what I was thinking,
>
> If it had a low turn out they might of considered it a failure and a waste
> of resources.  They days of getting high standard education online is
> getting closer and closer. Sure we could of all tried to learn this stuff
> ourselves by downloading books or reading tutorials like we normally do, put
> you'd end up putting it on the long finger. This sort of structure with set
> deadlines motivates you to keep going like physical onsite education.
>

> On Wed, Oct 19, 2011 at 2:52 PM, Ciaran Whyte <ciaran.wh...@gmail.com>wrote:
>
> > Awesome news, the bigger the count the more likely they'll be to do more
> > like it in the future
>
> > On Wed, Oct 19, 2011 at 2:47 PM, Brainfreeze (Lego Guru) <
> > br41nfr3...@gmail.com> wrote:
>
> >> No reason, just wondering.
>
> >> Apparently 46,000 actually did the AI homework. They confirmed.
>

> >> On Tue, Oct 18, 2011 at 5:10 PM, Ciaran Whyte <ciaran.wh...@gmail.com>wrote:
>
> >>> Public, as far as i know? why?
>
> >>> On Tue, Oct 18, 2011 at 5:09 PM, Brainfreeze (Lego Guru) <
> >>> br41nfr3...@gmail.com> wrote:
>
> >>>> Is this study group private or public?
>

> >>>> On Tue, Oct 18, 2011 at 5:02 PM, Ciaran Whyte <ciaran.wh...@gmail.com>wrote:
>
> >>>>>http://www.reddit.com/r/aiclass/comments/lfyq3/can_we_know_how_many_p...

>
> >>>>> Yeah I'd say you're right  -  I'm going to stick with it till the end
>
> >>>>> On Tue, Oct 18, 2011 at 5:00 PM, Brainfreeze (Lego Guru) <
> >>>>> br41nfr3...@gmail.com> wrote:
>
> >>>>>> How do you know how many people did the homeworks?
>
> >>>>>> And those views, if people watched the vid twice (I did with some),
> >>>>>> then that means even less people watched it.
>

> >>>>>> On Tue, Oct 18, 2011 at 4:56 PM, Ciaran Whyte <ciaran.wh...@gmail.com

> >>>>>>>>>>>> isaac.cleren...@gmail.com

[Brainfreeze (Lego Guru)]

I didn't have time to do the AI or ML assignments or quizzes last week, so does that count as dropping out? Maybe not in AI but I think in ML.

So busy, but that's made it even worse for me because I'm a week and a half behind, and since this is the Halloween weekend, unless I stay in there will be no way to catch up.

[Tony O'Connor]

Still in the game for now!  Though the AI videos are starting to increase in number and difficulty.  I thought the last set weren't too bad in AI but the result was quite disappointing, 50% I think.  Hopefully it's just numerical mistakes.

The ML class is going quite well, and it meshes very well with my econometrics class.

2011/10/27 Ciaran Whyte <ciaran...@gmail.com>