Roland Shen Reflections

[rolandshen]

Roland Shen Reflections Nov. 30, 2013

Thanksgiving week

Fernando Perez talks to the class about iPython Notebook. It was a very enlightening
talk and learning about the history of Python along with what the future holds for this
platform was inspirational. Definitely bring him back to talk to the class in future
semesters.

Roland Shen Reflections Nov. 23, 2013

Skype with Luen

Luen Skype notes

-error diagram 0-1 on axis for ETAs model conditional intensity tells you how
likely an earthquake will appear at any certain time

-high values of lambda on left hand, lower on right think of it as threshold at
which we turn on alarm work out a new tau pair for each new value of lambda how
do we evaluate nu and tau?

-tau is the proportion of time for the alarm being on if we turned it on when a
value of intensity exceeds lambda

-nu is the proportion of earthquakes that your alarm would miss

-code set up for the temporal, space-time ETAs and strawman version with
sensaible parameters

-spatial info is important for space-time ETAs and does better than just temporal
ETAs

-for a particular window length we can use interarrival times as the only thing
in prediction and then deduce function lambda from interarrival times
differentiate cdf to get pdf b/c absolute value of lambda doesnt matter for
error diagram

-go from ETAs to MDA adjusting window lengths to magnitude. same principle

-time and magnitude variables from 1984-2010 measured in days

-started 18 June 2004 to collect data

-be careful of using e^x or 10^x for conditional intensity

-doesnt make a practical difference in what type of magnitude we use (body-wave
and moment magnitudes)

-body-wave is more useful for larger magnitudes

-magnitude type matters in immediate aftermath of a big earthquake but otherwise
doesn't have much impact

-take distribution of lambda of taus and lambda of nus

Roland Shen Reflections Nov 16, 2013

Luen and Stark

1. The class has been in a lull since our presentation last week and groups seem
   to be confused on what the next step is. We haven't seen Aaron for a week and
   our class time has been a little unhelpful. Most of the work seems to be done
   outside of class and over the weekend. We're delving into the Luen model and
   magnitude optimization for alarm times.

2. My "aha!" moment of the week has not happened yet. We are currently meeting
   in Bechtel to work on the Luen model and hopefully we will figure something out
   in terms of optimization in our groups by Tuesday.

3. If I were to take a different approach this week, it would be to look at
   other groups' issues on the issue tracker. I don't know what other groups are
   working on for this week.

Roland Shen Reflections Nov 9, 2013

Class Presentations

1. Presenting in front of the class on Tuesday felt great. We were able to meet
   up at Bechtel as a group of presenters on Monday before class to discuss the
   status of each group and how we could move forward together as a cohesive unit.
   Our organization showed during our presentation as each presenter of each
   vertical group spoke for a few minutes in order of what the class needed to
   accomplish by December.

2. My "aha!" moment of the week occured Monday night before Tuesday's
   presentation. I understood what each group was doing and how it fit into the
   project as a whole. The big picture was now visible and not hidden.

3. If I were to take a different approach this week, it would be to set a meet
   up time with the other presenters sooner. We have already started talking about
   a weekly meeting for all presenters so this shouldn't be a problem moving
   forward.

Roland Shen Reflections Nov 2, 2013

Group Project 2

1. Working on the Luen/Stark paper with a group of all
   analyzers and no data curators or visualizers was not that tough of a task but
   there were still a few challenges. Going to office hours helped but there should
   be a sense of time when going to office hours. Learning about our goals and
   future tasks is helpful but not if it takes too much time to figure out.

2. My "aha!" moment of the week was talking to my analyzer groupmates about the
   MDA model and how we were going to approach the project through a critical path
   process. We also used SMART goals to narrow down a bigger end goal into smaller
   checkpoints so that the project could be more manageable.

3. If I were to take a different approach this week, it would be to spend less
   time in office hours. Although it is helpful to go in and ask questions, there
   is a certain point where diminishing returns lowers the efficiency of attending
   office hours.

Roland Shen Reflections Oct 26, 2013

Horizontal Groups

1. I believe that having a sign-in sheet would be beneficial in increasing class attendance.
   Although I feel like I learn more from office hours than in lecture, attending lecture gives
   the class a structure to base our work around. Attributing signing in for attendance to
   grades also gives students an added incentive. Not all students are here to grade grub
   but knowing that you stand somewhere gradewise also helps.

2. A'ha moments: Knowing the presenter's role within writing the paper. Presenters need
   to research articles related to the earthquake models and share these findings with their
   vertical groups.

3. If I were to change anything from this week, I would have attended office hours. I
   couldn't this weekend because I lost my phone and had to go home for the weekend
   to get a new one. Hopefully I didn't miss anything important. I also could have spent
   more time learning python.

Roland Shen Reflections Oct 19, 2013

Stark Lecture 2

1. I liked how we learned about new forms
   of programming in JSON and HTML 5, and how we can incorporate these into our
   learning of python. Stark's lecture also provided us with knowledge on what we
   are actually doing for this project. The facts about earthquake prediction
   helped a lot during our discussion and visualization of earthquake data for our
   group project. Also, sudo apt-get install python-dev is needed to install pandas
   and is a useful line of code.

2. I had many "a-ha!" moments this week, especially during office hours. I
   helped my group members and classmates in other groups pull request their
   reflections and by teaching them how to maneuver through github, I had a better
   understanding of how to go through github myself. My groupmate Khoa also found a
   nice python package called folium and we used that in our project to show
   earthquake positions (longitude, latitude) on a world map. Teresa made a great
   point about collecting Stark notes on a github page as opposed to a google doc.
   Just create a stat157 repository for Stark notes and then have everyone fork and
   pull request individual notes onto the repository so that everyone can have easy
   access without having to click on a google doc link.

3. If I were to change anything from this week, it would be to focus more at
   the beginning of the Stark lecture. I learned a lot from his presentation but
   more towards the end when he started to bust out all the models in formula form.
   Paying attention at the beginning of the lecture would probably provide more
   insight into the "why" instead of the "how."

Stark Notes below:

poisson, gamma renewal models dont fit stoichastic seismic
data bigger events have smaller windows, shorter event times

automatic alarm: after every event with M > mu start an alarm of duration tau no
free parameters

magnitude-dependent automatic alarm (MDA) afte every event with M > mu start an
alarm of duration tau*u^M 1 free parameter ^none of this stuff is physics based

For both adjust fraction of time covered by alarms through tau

ETAS model in Luen PhD Dissertation

project in a nutshell, get a model that is better than ETAS at all points

Gutenberg-Richter law for magnitudes
Omori-Utsu law for aftershocks

Roland Shen Reflections Oct 12, 2013

Group Project 1

1. The group project was a good change of pace from all that we have done so
   far. We were thrown into the fire and had to collaborate with our team of four
   to survive on such short notice. This forced us as students to focus on this
   class and put it as a priority over other classes and I learned more about
   iPython Notebook through this exercise than through any lecture beforehand.

2. My "aha!" moment of the week was when I figured out how to run iPython
   Notebook from my VM. I also learned about background and foreground processes
   from my technical lead (who also happened to be in my group) so that I could run
   iPython Notebook in the background while working on other things in the
   foreground. I never knew such a thing existed and I hope to learn more as the
   class progresses.

3. If I were to take a different approach this week, it would be to start on the
   group project sooner. We talked over email a little bit after Thursday but we
   did not meet up as a group until Sunday at Aaron's office hours, which was
   poorly planned. Starting sooner would have allowed for us to sooner encounter
   the setbacks we had and we could have produced a better end product. I know
   perfection is the enemy but seeing the other groups' projects made me feel that
   ours was a bit lackluster.

Roland Shen Reflections Oct 4, 2013

Stark Notes Included

Measure earthquake as a stochastic process using poisson process and memoryless
properties to predice earthquakes estimate how big earthquakes are using moments
in physics earthquake probability is a metaphor claims that events occur as if
in a casino game, why not like terrorism instead?
La quila earthquake in italy 2010
Urn model is just an assumption, just because there is a rate that there is a
probability

Analogy of earthquake prediction to weather prediction obviously you would
predict differently based on past events like if it rains today, it is more
likely to rain tomorrow than it would in a summer. If an earthquake happens,
aftershocks will be more likely to happen.

**Assuming that something is random is a big assumption to make

I think that this week was a vast improvement from last week.
I had a hard time setting up virtual machine because I forgot my password on the
VM that I had already downloaded. That was easily fixed though, along with the
local port.

An aha moment that I had this week was when Professor Stark talked about how we
should treat earthquake occurances. Now that we know how earthquake prediction
works and the consequences of those who try to predict earthquakes and fail
really puts this class into perspective when it comes to reproducibility and
statistical analysis.

We have a goal that we can work toward in our groups and we also now have a
schedule that we can follow. I am looking forward to next week's presentations
on Tuesday and working with my group over the weekend.

Roland Shen Reflections Sept 28, 2013

What are your thoughts on reproducibility?

I think reproducibility is a great idea for people who are interested in
research and making an impact within the academic community. It allows for
others to understand your work step by step and to be able to create the same
results from your findings. Although I might not be interested in doing
research, I believe that reproduceability is important even within the
workplace. If I had to complete a project within a company, I would like to be
able to reproduce the same quality work later if a similar project shows up
again. I wouldn't want my first successful project to be a product of luck and I
would want to be absolutely sure that if I followed the same steps from the
previous project, then the current project would yield the same results. The
strongest justification for the opposite view is that some measurements in data
are variable based, that even though all the steps are carefully recorded and
followed, different results can appear. For such experiments or processes, it is
important to say at the beginning of a process that there will be variations in
findings but only because that is the nature of the research at hand.

Roland Shen Reflections Sept 21, 2013

We only had two classes with guest lecturers this week. I don't know if there
are any assignments related to these lectures.

I still have not felt an "A-ha!" moment in this course but hopefully when the
class starts to pick up next Tuesday there will be more opportunities to learn
new material.

Again there was no new material this week and as such I wouldn't have done
anything different. I'm glad that there were guest speakers this week.

Roland Shen Reflections Sept 17, 2013

I was unaware that there would be a second reflection this week. I feel that our
class discussion in Thursday's section covered the main points that the course
should improve on. As far as communication between staff and student goes, I was
unsure whether we should submit anything if we did not do anything programming
related in class for the week. I also did not know if we should attach the
second reflection to the first or to create a new repository. Again, not having
a set assignment schedule can be confusing but we went over this in class.

I still have not felt an "A-ha!" moment in this course but hopefully when the
class starts to pick up next Tuesday there will be more opportunities to learn
new material.

Again there was no new material this week and as such I wouldn't have done
anything different. I'm glad that the GSIs are hosting office hours from 9am-
noon on Friday so that everyone can be on the same page.

Roland Shen Reflections Sept 7, 2013

1. I had a hard time setting up my virtual machine until group leaders in class
   helped me out. Following instructions for downloading ipython and IRC was also
   confusing. I understand what I should have done now that I have everything set
   up, but I feel that the instructions could have been clearer. There was a lot of
   computer lingo that I couldn't process espectially after I had installed the VM
   VirtualBox. I didn't know how to do anything in homework-1 after the
   installation.

I am also trying to understand how to use github through the octocat tutorial
since github still feels foreign to me. I need to get used to knowing what
pulling a repository means or pushing something to someone else, as well as how
to navigate through the website efficiently.

2. I have not yet had an "aha!" moment but maybe that is because I haven't spent
   enough time learning github. I have tried to use Git Shell for this assignment
   but that hasn't gone well so far.

3. If I were to take a different approach to this past week, I would have spent
   more time in becoming acquainted with github. I am not a very fast learner,
   especially when it comes to programming, so spending more time definitely would
   have helped. For other students, I suggest using IRC a lot. Whenever you have a
   problem, someone else would probably also have the same problem and maybe
   figured out a way to solve it by the time you ask. I didn't know what to do for
   this week's reflection so I went onto IRC and asked. Someone told me
   instructions to come to this page and post.

[aculich]

Can you format this markdown file according to the basic conventions here:

https://github.com/carwin/markdown-styleguide#basic-conventions-for-markdown-files

In particular, 80 character maximum for the line length. And break things into multiple
paragraphs.

That will make it easier for me to make comments using the github line commenting
system.

Also, you probably don't want to format each line by hand. If you use a text editor
like http://www.sublimetext.com/3 you will be able to have it format to 80 characters
for you so you don't have to do it by hand.

Once you re-format it, then submit a new pull request.