hw1 - class materials don't help

Evgeny

unread,

Feb 6, 2013, 3:34:56 PM2/6/13

to 10-701-spri...@googlegroups.com

In my opinion, lectures explain little about the homework, like implementation of Naive Bayes in discrete vs continuous cases. You can see that by the number of questions here (like, what is the training/testing error). Surprisingly, had to find more or less all the implementation techniques in wikipedia (with examples and even numbers there)

Jingkun Gao

unread,

Feb 6, 2013, 3:44:28 PM2/6/13

to 10-701-spri...@googlegroups.com

Agreed. I think it will be better if there is a high rho between the homework and the lecture. For example, the implementation of the algorithms in lecture to the homework will also help further understand the algorithms.

Madalina Fiterau

unread,

Feb 6, 2013, 4:14:00 PM2/6/13

to Jingkun Gao, 10-701-spri...@googlegroups.com

Hi everyone,

The purpose of the lecture is to explain key-concepts in ML and to give an overview of how the algorithms work (and mathematically show why they work).

Implementation details are really out of scope for the lecture, but we can cover some in the recitation.

For the record, none of the ML/CS classes that I took detailed algorithm implementation in the lecture, so I'd say the current state is pretty standard for ML-related courses. Depending on your background you might or might not be used to this.

We're doing our best to answer your questions, and we'll be more explicit in future implementation requirements.

Thanks.

Ina

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-201...@googlegroups.com.
To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

milad memarzadeh

unread,

Feb 6, 2013, 4:16:22 PM2/6/13

to Jingkun Gao, 10-701-spri...@googlegroups.com

Agreed. I think the lectures are too much focused on the theoretic stuff behind the learning methods which could be materials for further reading for those who are interested. It's better to focus more on the implementation details of different methods than the theory behind it.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-201...@googlegroups.com.
To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
Milad Memarzadeh, M.Sc.
Doctoral Candidate, Advanced Infrastructure Systems
Department of Civil and Environmental Engineering
Carnegie Mellon University

milad memarzadeh

unread,

Feb 6, 2013, 4:18:27 PM2/6/13

to Madalina Fiterau, Jingkun Gao, 10-701-spri...@googlegroups.com

Inna,

By focusing more on the implementation details of ML techniques I refer to the class by Andrew NG from Stanford. I guess that course is also a ML/CS course which I find much more useful.

Thanks,

Milad

Se-Joon

unread,

Feb 6, 2013, 4:25:06 PM2/6/13

to 10-701-spri...@googlegroups.com

I have also found previous CMU ML lectures useful. For example, for implementing Naive Bayes:

http://www.cs.cmu.edu/~guestrin/Class/10701-S06/Slides/naivebayes-logisticregression.pdf

On Wednesday, February 6, 2013 3:34:56 PM UTC-5, Evgeny wrote:

Matineh Eybpoosh

unread,

Feb 6, 2013, 4:39:06 PM2/6/13

to Se-Joon, 10-701-spri...@googlegroups.com

I agree too.

I guess understanding the mathematical theory behind the algorithms is important for their implementations, however, without some tangible examples it is not much possible to understand the theory too.

Sometimes, in lectures, I feel like I am lost at the middle of bunch of complex mathematical proofs and symbols that I am not quite sure how to picture them in a real scenario.

Regards,

Matineh

Abhinav Jauhri

unread,

Feb 6, 2013, 5:05:53 PM2/6/13

to Matineh Eybpoosh, 10-701-spring-2013-cmu, Se-Joon

I think if we are provided with lecture notes then it would be very helpful in revising/understanding a concept/proof taught in class, preparing for exams, and even to maintain some consistency with notations throughout the course. Something similar to these - http://cs229.stanford.edu/materials.html.

I am, personally not good at making notes and listening at the same time in class.

--
Abhinav Jauhri

Jerry Vinokurov

unread,

Feb 6, 2013, 5:27:12 PM2/6/13

to 10-701-spring-2013-cmu

My sense is that the lectures range pretty widely and serve mostly to introduce the concepts. There's just too much material to go into in-depth proofs in the lecture, but it would be nice if we could get some references in the slides or elsewhere to either more detailed discussions (e.g. books, articles) or implementation examples.

Jerry

http://www.google.com/profiles/grapesmoker

Barnabas Poczos

unread,

Feb 6, 2013, 5:35:59 PM2/6/13

to Jingkun Gao, 10-701-spri...@googlegroups.com

Dear Students,

Many thanks for your feedback!

I think one difficulty is that there are 145 students attending this
class and you are all very different:

Some of you found the homework too easy (including the Naive Bayes
implementation), while others told us that the homework was too
difficult.

It is quite challenging to find a good balance that makes everyone
happy. Please be patient, and help us with sending more feedbacks!

What I can recommend is

- Visit the office hours if you have questions. The TAs and the
instructors will be very happy to help you and explain those parts
that you don't understand.
For example, if you have questions about implementations of naive
Bayes, I will be happy to discuss this with you every week Thursday
10:00-12:00.
So far only one person came to my office to tell me that he didn't
understand the Naive Bayes method.

- Ask more questions during lectures. If you don't ask, we don't know
what is not clear.

- Visit the recitations

- On the course website we also provide other materials that can help
further understand the lectures. If it is still not enough, we can
send you more papers, tutorials, surveys, and books that can help you;
just let us know!

- At the beginning of each lecture we always ask if there is any
question about the previous material. Please ask if you have
questions!

- About the lecture notes: I agree, it would be very helpful indeed!
Since the course is video recorded I think it could be relatively easy
for you to self-organize this and write notes about the lectures.
There are 145 students for about 25 lectures, so in theory it should
be easy. Alex and I will be very happy to read your notes and correct
it if needed.

- Finally, please keep in mind that this is a graduate course. I
understand that some of you want to learn more about implementation
details, but we also have to think about those students who are
interested in theory, want to know why these algorithms work, what
their theoretical properties are, and what the open questions are
these days in machine learning.

Suggestions are always welcome! I really appreciate them.

Thanks again,
Barnabas

Enze

unread,

Feb 6, 2013, 5:46:49 PM2/6/13

to 10-701-spri...@googlegroups.com

Agree. I also noticed the schedule is very different from previous 10701s and goes much deeper into mathematics. I attended some of the lectures last semester. It was hard but not to this level. e.g. they used one lecture on MLE/MAP and another on NB, with more concrete examples. http://www.cs.cmu.edu/~epxing/Class/10701/lecture.html

Madalina Fiterau

unread,

Feb 6, 2013, 5:53:52 PM2/6/13

to milad memarzadeh, Jingkun Gao, 10-701-spri...@googlegroups.com

Hi Milad,

What I meant was that it's not the point of the lecture to explain how to implement specific things given to you as homework (see complaint of the first two students). An example: you will not be shown how to implement a constrained optimization solver though SVM uses it. It's the other way around: the point of the homework is (1) to ensure that you understood what was shown in class (2) to ensure the course is detailed enough to be useful for practical problems.

It seems though the problem is more deeply rooted. It looks like you guys aren't getting enough of an idea of how these methods are supposed to work with the actual data. That can be fixed.

We'll cover examples in the recitation, so please help by answering the recitation topic saying which methods you didn't understand or don't see how they would be applied in practice.

Some problems will only arise when you're actually faced with a task (see discrete vs continuous variables). That's part of the learning process for everyone: students, instructors and TAs. We're doing ML for a living so maybe we're taking some things for granted when we shouldn't - like test/train error.

So everyone: thanks for pointing out the issues you're facing with the class. It will help us improve.

Ina

milad memarzadeh

unread,

Feb 6, 2013, 5:58:55 PM2/6/13

to Madalina Fiterau, Jingkun Gao, 10-701-spri...@googlegroups.com

Thanks a lot Ina for your email and clarification. I agree with you to some point but as I remember the instructors mentioned in the first class that the slides tagged with red color means you can sleep and so far 70% (and even more) of each lecture slides were tagged red. What I mean is that it should be focused more on yellow and green tagged slides than red ones.

Thanks for your email and I have already suggested my concerns in recitation lectures.

Best,

Milad

Alex Smola

unread,

Feb 6, 2013, 6:09:08 PM2/6/13

to Matineh Eybpoosh, Se-Joon, 10-701-spri...@googlegroups.com

guys,

i find the discussion quite surprising - today i had the class vote on whether i should cover a much more theoretical subject (regularization theory) OR something very practical that's running on spam filters in a major company. the audience voted (the voting audience, that is) with overwhelming majority for theory.

so - if you want practical things, you need to participate and make yourself heard. replying on the forums what you would have _wanted_ rather than telling what you want in class isn't very useful.

as for the homework - i had several complaints that the homework was too easy. now i'm hearing that you feel at a loss doing it. i'm confused. this time's homework will have a more diverse range of difficulties.

as for the programming assignments, for discrete variables we discussed laplace smoothing. for continuous variables we discussed kernel smoothing. this is more than plenty for the purpose of naive bayes.

i'm happy to adjust the topics of the class, slides and homework according to what is desired. but i cannot meet a goal and its opposite.

cheers,

alex

--

''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+

| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |

+-------------------------\ (----( )-----------------------------+

\_) ) /

(_/

Humphrey Hu

unread,

Feb 7, 2013, 12:10:43 AM2/7/13

to Alex Smola, Matineh Eybpoosh, Se-Joon, 10-701-spri...@googlegroups.com

Hi Alex,

Thank you for being so receptive to all our feedback. I understand it is difficult to balance the needs of a diverse class, but I have yet another view on the course so far.

First, I disagree with some of the previous sentiments that lecture should completely prepare us to do the homework. I have no qualms with looking up implementation details on my own if I have a strong grasp of the underlying theory.

However, try as I may, I haven't been learning the theory because I have not been able to follow lecture at all since day 2 or 3. This is mainly due to my unfamiliarity with ML terms even as basic as "loss" and "risk" that are introduced without definition in the slides. Many mathematical terms in equations are also used without introduction, such as /nu at the beginning of lecture today (I was intently listening to an explanation I couldn't follow, so I can't even specify the topic). Looking these up after lecture helps, but it means that lecture itself is overall useless to me.

Maybe I am alone in my confusion. If so, I will turn to office hours and background reading, but if others feel similarly, I think a few seconds devoted to defining terms per slide would be extremely helpful.

Thanks for reading,

Humphrey

Message has been deleted

Karthik

unread,

Feb 7, 2013, 12:44:01 AM2/7/13

to 10-701-spri...@googlegroups.com, Alex Smola, Matineh Eybpoosh, Se-Joon

Hi Alex,

Firstly, thanks to all the instructors for being responsive to comments on the forum! I just want to add a bit to Humphrey's comments.

I agree with Humphrey on the point about preparation for homework - while I wouldn't say I found the homework too easy, I didn't think that the lectures left me unprepared for it. I think it's perfectly fine to focus more on concepts in lecture with a little bit of application, which has definitely been the case so far. In fact I really appreciate the effort that has been taken to put in real life examples like spam filtering etc into pretty much every lecture :)

I have had some exposure to these concepts in previous courses, but I am still finding the pace of lectures hard to keep up with. Personally, I would prefer if each lecture covered a little less material, and went slower and more in depth into the basics and motivation for each big idea at the beginning of lecture. I think spending some extra time on the first few slides of each lecture would be very helpful

So far, at some point in the lecture I tend to get lost in the math without a clear understanding of what the motivation behind the math is. A little more time spent on problem set up (including defining variables and parameters) would be fantastic.

I hope I am not the only one feeling this way.

Thanks!

Karthik

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

Barnabas Poczos

unread,

Feb 7, 2013, 1:02:08 AM2/7/13

to Karthik, 10-701-spri...@googlegroups.com, Alex Smola, Matineh Eybpoosh, Se-Joon

Hello Everyone,

If you collect and send me which slides are difficult to follow, which
notations have not been defined on the slides, which slides are
confusing, etc..., then I'm happy to redo and extend my slides with
this extra information.

Many thanks again for your constructive suggestions!

Best,
Barnabas

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>>
>>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>>> (YouTube playlist)
>>>> ---
>>>> You received this message because you are subscribed to the Google
>>>> Groups "10-701 Spring 2013 CMU" group.
>>>> To unsubscribe from this group and stop receiving emails from it, send

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>>
>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>
>>> --
>>> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
>>> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
>>> (YouTube playlist)
>>> ---
>>> You received this message because you are subscribed to the Google Groups
>>> "10-701 Spring 2013 CMU" group.
>>> To unsubscribe from this group and stop receiving emails from it, send an

>>> email to 10-701-spring-201...@googlegroups.com.

>>> To post to this group, send email to 10-701-spri...@googlegroups.com.
>>>
>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>
>>>
>>
>>
> --
> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
> (YouTube playlist)
> ---
> You received this message because you are subscribed to the Google Groups
> "10-701 Spring 2013 CMU" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to 10-701-spring-201...@googlegroups.com.

Ruikun Luo

unread,

Feb 7, 2013, 1:13:53 AM2/7/13

to 10-701-spri...@googlegroups.com, Matineh Eybpoosh, Se-Joon, al...@smola.org

Hi Alex,

I agree with Humphrey so much. I think guys like us has no machine learning background. The math is not the problem, but we need more background knowledge to understand what behind the equations. I think it's better to introduce the notations and definitions first. And I think if we could get some materials about some basic knowledge that will mentioned in the next lecture, it will help us to keep with the lecture.

Thank you for your understand.

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

zhuoc

unread,

Feb 7, 2013, 10:00:14 AM2/7/13

to 10-701-spri...@googlegroups.com, Matineh Eybpoosh, Se-Joon, al...@smola.org

Hi,

I agree with Humphrey and guys above. And, same as them, I really appreciate the instructors' responsive replies to this forum.

Sometimes I feel I can't find the _connections_ between different parts of the lecture. One example would be when learning perceptron, we were given equations like f (x) = (<w, x> + b), and I thought "okay, we are now trying to solve a regression...", and it was not. I might have missed something in the lecture, but it did take me quite a while to figure out we were doing classification based on whether f(x) is larger or smaller than 0. (or am I still wrong now?...) So this is what I mean by "connection", the algorithm after that is not that hard for me. So I hope that before introducing a new algorithm, what problem this is going to solve can be clearly stated (just one sentence). Is it classification/regression, can the features be continuous, discrete or categorical?

Thanks,

Zhuo

Matineh Eybpoosh

unread,

Feb 7, 2013, 10:45:01 AM2/7/13

to Karthik, 10-701-spri...@googlegroups.com, Alex Smola, Se-Joon

Hi Alex,

I just wanted to vote for the last two comments from Karthik and Humphrey. I guess what they explained is exactly what is happening to me too.

I also agree that specific issues addressed in the homework should be our job to search for and figure out...actually, I believe we would better learn in this way.

However, as is mentioned in the last two emails, spending some time at each lecture on "explaining why we are learning this method and the transition from the past topics, defining the concepts, symbols, and the goals of the methods" in plain English would help students out of CS, ML department to trace the rest of the lecture better, and take the most benefit out of your efforts.

Thanks again for providing us with such a platform to share our opinions.

Regards,

Matineh

On Thu, Feb 7, 2013 at 12:41 AM, Karthik <karthik...@gmail.com> wrote:

Hi Alex,
Firstly, thanks to all the instructors for being responsive to comments on the forum!

I just want to add a bit to Humphrey's comments.

I agree with Humphrey on the point about preparation for homework - while I wouldn't say I found the homework too easy, I didn't think that the lectures left me unprepared for it. I think it's perfectly fine to focus more on concepts in lecture with a little bit of application, which has definitely been the case so far. In fact I really appreciate the effort that has been taken to put in real life examples like spam filtering etc into pretty much every lecture :)

I have had some exposure to these concepts in previous courses, but I am still finding the pace of lectures hard to keep up with. Personally, I would prefer if each lecture covered a little less material, and went slower and more in depth into the basics and motivation for each big idea at the beginning of lecture. I think spending some extra time on the first few slides of each lecture would be very helpful

So far, at some point in the lecture I tend to get lost in the math without a clear understanding of what the motivation behind the math is. A little more time spent on problem set up (including defining variables and parameters) would be fantastic.

I hope I am not the only one feeling this way.

Thanks!
Karthik

On Thursday, February 7, 2013 12:10:43 AM UTC-5, Humphrey Hu wrote:

Humphrey

Matineh

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

--
''~``

( o o )

+----------------------.oooO--(_)--Oooo.---------------------------+
| Prof. Alexander J. Smola http://alex.smola.org |

| 5000 Forbes Ave phone: (+1) 408-759-1044 |

| Gates Hillman Center 8002 Carnegie Mellon University |

| Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
+-------------------------\ (----( )-----------------------------+

\_) ) /
(_/

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

--
Matineh Eybpoosh

------

PhD Student

Civil and Environmental Engineering

Carnegie Mellon University

5000 Forbes Ave, Pittsburgh, PA 15213

Tel: 412-251-3853

www.matineheybpoosh.com

Alex Smola

unread,

Feb 10, 2013, 7:33:34 PM2/10/13

to Matineh Eybpoosh, Karthik, 10-701-spri...@googlegroups.com, Alex Smola, Se-Joon, Barnabas Poczos

hi matineh,

ok, we'll try adding more motivation. i assumed that, e.g. the discussion for novelty detection (there were three slides only on motivation) plus two slides of examples for ocr would have been sufficient. but we should probably add even more.

it might help if you guys could let me know what the machine learning problems in your phd research are. then i can refer to them some more. most of my motivations are internet related. this is simply the case since i assume that this is a medium that we all use. plus i've spent the past five years working for internet companies.

cheers,

alex

Matineh

Matineh Eybpoosh

------
PhD Student
Civil and Environmental Engineering
Carnegie Mellon University
5000 Forbes Ave, Pittsburgh, PA 15213

Tel: 412-251-3853
www.matineheybpoosh.com

--

http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-201...@googlegroups.com.

To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Victor Hwang

unread,

Feb 14, 2013, 2:30:06 PM2/14/13

to 10-701-spri...@googlegroups.com, Matineh Eybpoosh, Karthik, Alex Smola, Se-Joon, Barnabas Poczos

Hi Alex and Barnabas,

I'd also like to thank you for being so receptive to suggestions. This may be the first class I've been in where the professors were so actively trying to get feedback, so it's much appreciated.

I would also like to mirror Karthik and Humphrey's sentiment on getting lost during the lecture. The amount of detail on each slide makes it difficult to follow for the full 80 minutes. However, in the spirit of machine learning, I'll try to quantify things a bit more instead of echoing what's already been said.

I took two lectures (Alex's lecture 5 on kernels and Barnabas' lecture 8 on theory) and did a few counts. I also included counts of 2 of Tom Mitchell's lectures on SVMs for comparison because I find his lectures very helpful.

Alex's lecture: 55 math-related slides (~114 total slides) containing 161 equations* over 160 minutes (2 lectures). Total number of words: 1419.

Barnabas' lecture: 34 math-slides (60 total slides) containing 150 equations* over 160 minutes. Total number of words: 567.

Mitchell's lecture: 33 math-slides (50 total slides) containing 109 equations* over 160 minutes. Total number of words: 656

* - what I considered "non-trivial" equations, or equations with more than 3 symbols. Usually containing equal signs. Difficulty of equations not considered...though it does make a big difference. Mitchell tends to have the same equations over and over again, so there is a bit of double counting.

This gives (conservatively, because I make the assumption that all minutes are spent explaining equations)

Alex: 2.92 eq/math-related slide @ .99 min/equation

Barnabas: 4.4 eq/math-related slide @ 1.01 min/equation

Mitchell: 3.3 eq/math-related slides @ 1.46 min/equation

I understand that there's a lot of material to get through, and it must be very difficult to have to cram them into 90 minutes. Machine learning IS mathematics and it's impossible to boil down too much. However, I think you will be able to get more participation if lectures focused more on the higher level ideas that can be explored in a depth-first manner. Perhaps separating out the heavy mathematical details from the lecture for later review could help with this.

Again, thank you for allowing us to have this open discussion.

[Mitchell's lecture slides. His videos are also available on his site:

http://www.cs.cmu.edu/~tom/10701_sp11/slides/Kernels_SVM_04_7_2011-ann.pdf

http://www.cs.cmu.edu/~tom/10701_sp11/slides/Kernels_SVM2_04_12_2011-ann.pdf]

Helen

unread,

Feb 14, 2013, 8:21:52 PM2/14/13

to 10-701-spri...@googlegroups.com

Hi Alex,
I'm from CS and my research is related with machine learning application. So my expectation is to have a deeper understanding of some algorithm, which can help me to analyze and solve the problem I came across in future.

One suggestion of the course content is that I hope you can pay more attention to the notation.
When I refer to the slides after class, I can often find something unclear, like what does some symbol in a formula mean. Though I might catch up the key idea of most of the stuff, I can't figure out stuff like this! I really hope you can double-check the slides (maybe do it with your TAs too) to find out whether there is something not clear enough.

Also, I hope to learn some theory in the class instead of only application (I think this often benefits from doing something in their field by applying ML instead of learning in class). This would be helpful to guide me in solving and analyzing problems.

However, since I'm not in theory track, I don't want to spend too much time in proof. So one suggestion is that give some problems related with theory in homework, but give enough hints and background so that it's feasible.

Thanks.

On Wednesday, February 6, 2013 3:34:56 PM UTC-5, Evgeny wrote:

Alex Smola

unread,

Feb 17, 2013, 9:12:51 PM2/17/13

to Victor Hwang, 10-701-spri...@googlegroups.com, Matineh Eybpoosh, Karthik, Alex Smola, Se-Joon, Barnabas Poczos

hi victor,

excellent analysis. i never thought of parametrizing my slides in terms of epm (equations per minute) but this is probably really useful. will try to stay below 1 epm.

so - here's why there were so many equations in it. i had requests that i go through the derivation in a lot of detail. a lot of detail for svms means actually stepping through every line of the wolfe dual derivation. that means writing out the primal. in detail. writing out the lagrange function. computing derivatives. plugging it in and getting the dual.

the intention was to help with understanding by providing detail. plus there were several supplementary slides which contained derivations if anyone wanted them to look up after class.

so - if you find it easier, i can omit these equations. and focus on the main parts. but then you won't know where the math came from. i'm happy to do either of the two (actually the latter is much less work for me, so i really don't mind doing less).

let me know what you guys prefer and i'll adjust accordingly.

cheers,

alex

--
http://alex.smola.org/teaching/cmu2013-10-701 (course website)
http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9 (YouTube playlist)
---
You received this message because you are subscribed to the Google Groups "10-701 Spring 2013 CMU" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-201...@googlegroups.com.
To post to this group, send email to 10-701-spri...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

Manzil Zaheer

unread,

Feb 18, 2013, 12:09:58 PM2/18/13

to 10-701-spri...@googlegroups.com, Victor Hwang, Matineh Eybpoosh, Karthik, Alex Smola, Se-Joon, Barnabas Poczos

Hello Alex,

I would vote for keeping the things as it is, i.e. providing details including several supplementary slides which contained derivations.

Thanking you,

Manzil

To unsubscribe from this group and stop receiving emails from it, send an email to 10-701-spring-2013-cmu+unsub...@googlegroups.com.
To post to this group, send email to 10-701-spring-2013-cmu@googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

Bryan

unread,

Feb 18, 2013, 5:51:34 PM2/18/13

to 10-701-spri...@googlegroups.com, Victor Hwang, Matineh Eybpoosh, Karthik, Alex Smola, Se-Joon, Barnabas Poczos

I definitely agree with Victor on this. The slides contain so much information, that coupled with the verbal lecture it is nearly impossible to take everything in. So without the big picture (a few key concepts and equations), all the details don't add up to much importance.

Bryan

Barnabas Poczos

unread,

Feb 18, 2013, 6:17:07 PM2/18/13

to Bryan, 10-701-spri...@googlegroups.com, Victor Hwang, Matineh Eybpoosh, Karthik, Alex Smola, Se-Joon

If you have questions on the slides, I'm happy to explain them to you.
Please visit me in my office hours, and let me know which slides cause
problems or what parts you didn't understand.

Thanks,
Barnabas

>>>> an email to 10-701-spring-201...@googlegroups.com.

>>>> To post to this group, send email to

>>>> 10-701-spri...@googlegroups.com.

>>>> For more options, visit https://groups.google.com/groups/opt_out.
>>>>
>>>>
>>>
>>>
>>>
>>> --
>>> ''~``
>>> ( o o )
>>> +----------------------.oooO--(_)--Oooo.---------------------------+
>>> | Prof. Alexander J. Smola http://alex.smola.org |
>>> | 5000 Forbes Ave phone: (+1) 408-759-1044 |
>>> | Gates Hillman Center 8002 Carnegie Mellon University |
>>> | Pittsburgh 15213 PA ( ) Oooo. Machine Learning Department |
>>> +-------------------------\ (----( )-----------------------------+
>>> \_) ) /
>>> (_/
>>>

> --
> http://alex.smola.org/teaching/cmu2013-10-701 (course website)
> http://www.youtube.com/playlist?list=PLZSO_6-bSqHQmMKwWVvYwKreGu4b4kMU9
> (YouTube playlist)
> ---
> You received this message because you are subscribed to the Google Groups
> "10-701 Spring 2013 CMU" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to 10-701-spring-201...@googlegroups.com.

> To post to this group, send email to

> 10-701-spri...@googlegroups.com.

Reply all

Reply to author

Forward