I have finally finished my first sprint of the Personal PhD. I finished it later than my target of Dec 25th, but hey, at least I finished what I set out to do. (Here is my first sprint announcement post).

What got done

I finished watching all videos and in-course exercises of Udacity’s Introduction to Inferential Statistics. I am glad I did! After having never taken statistics in school this was very useful. Now I understand what p-values mean and how to utilize t-tests, z-score, confidence intervals, hypothesis testing, ANOVA, r^2 etc.

I read the book The Theory That Would Not Die: How Bayes’ Rule Cracked the Enigma Code, Hunted Down Russian Submarines, and Emerged Triumphant from Two Centuries of Controversy by Sharon Bertsch McGrayne and it was a fascinating read. I now have a more intuitive understanding of Bayes rule and it was fascinating to read about the history of how it was applied.

I also listened to all episodes of Talking Machines podcast. I didn’t necessarily understand everything, but it was still interesting (http://www.thetalkingmachines.com/).

The format of the sprint

I like this 6 week length sprint better than 3 month length quarter. It is shorter so there is less chance for life to get in the way. The deadline is sooner so if life does get in the way, I have to summon the energy to finish stuff sooner. As a result less debt accumulates, and because of that it is easier to finish and get to the end.

On note-taking

Throughout school and college I have always taken notes with pen and paper. So far in my Personal PhD I haven’t. I have taken few notes in a Google doc, but it is not nearly as comprehensive as the notes I used to take.

I am starting to think that not taking notes or taking way less notes digitally is not as productive for learning as taking more physical notes, at least for me.

I have known for a while that I am visual learner.  I remember in school when we had to learn poems by heart and then recite them, I would write them down on a piece of paper and then take the paper with me everywhere and learn the poem from it. When I had to recite, in my mind, in front of me there would be a visual picture of the piece of paper with the poem and I would basically read it out from the picture in my mind.

If you give me verbal directions, left, right, straight, left at the light, there is no chance I will remember them. But a quick glance at the map with directions and I will be able to find my way.

The video presentations online MOOC like Udacity’s work well for me (way better than just listening to a lecture), but I think I am missing on the note taking part. If I write down a formula by hand in my notes I get to see it again and again as I continue taking notes and referring back to it, thus engraving a visual picture in my mind. Also, without physical notetaking I don’t have a summary cheat-sheet of the main concepts. So I think for next courses I will try to do physical notetaking with a pen and paper.

How much time did I spend?

Dominique once asked me how much time do I spend on this project. For this sprint I kept track. Studying on Udacity took 16.7 hours, by far the vast majority was in the last week between Christmas and New Year’s, when I was doing about an hour a day to finish it before New Year’s.

There are 41 episodes in the podcast and total length almost 27 hours. That seems quite a lot, but it didn’t felt I was spending extra time on them. I found the time while doing other activities or while waiting: I listed while running, working out, walking, waiting, on airplane, at dentist’s etc.

Don’t know how long it took to read the book, but it felt like quite a long read 🙂

The prep week of the first Personal PhD sprint is done. I have chosen my curriculum. I have received few questions on how I pick courses among sea of so many options, so I thought I would share my process in this case.

In general my philosophy is, if there are many different course options, it doesn’t really matter which specific course you pick to go through content wise. From any of them, you will learn more than if you are being paralyzed from too many choices and don’t make a decision at all and don’t study. This is one of the reasons I have an explicit prep time built into my structure so that 1) I give myself time to play around with different options 2) have a deadline by which I have to pick a material. I feel that gives a balance of having somewhat informed decision, but not get bogged down in endless research.

Previously I took Udacity’s Intro to Descriptive Statistics course, now it’s time to learn about Inferential Statistics.

I did some brief Googling on online statistics courses, incorporated coworker recommendations, and looked at next course in sequence for Udacity and picked three options to check out in detail. They were the following:

1. edX Introduction to Statistics by University of California Berkeley.
2. Probability and Statistics on Stanford Online platform.
3. Udacity’s Intro to Inferential Statistics course.

I spent roughly an hour on each and went through the first few lessons.

For me personally, what ended up being the main criteria in this case for picking one course over another was how mobile-friendly it was.

I spend a lot of time with my butt in a seat, be it plane, train or shuttle. I want to use that time productively, and one the ways how I spend it is studying. So bite-sized lessons being available on the phone is important.

If you always study at certain time at a desk with a computer then this totally doesn’t apply to you.

Here are my thoughts on each course.

EdX Berkeley course:

• Consists mostly of video lessons with the professor speaking quite slowly. On the desktop it is possible to increase the speed of the video, but not on the mobile site.
• The video screen is small by default on mobile and on each video needs to be expanded bigger; in the expanded mode there is no way to advance to the next video. So unzoom, advance, zoom.. too clumsy for a mobile interface.
• There is additional reading and exercises on a different webpage that is not really integrated with the course. It is not always clear which exact section need to be read and in which order compared to the videos. Too many micro-decisions cause decision fatigue and therefore there is less willpower left for the actual studying. Also it is very clumsy on a smartphone to jump back and forth between different webpages.

Stanford Online Platform also uses EdX to power its courses, however their Statistics and Probability course is actually very heavy on text materials, not just video so it wasn’t that painful to not be able to speed up the occasional video on the phone. It had an excellent introduction on meta-cognition to be a more effective learner in MOOC setting, which I thought was a very nice touch.

Udacity has a dedicated mobile app which is quite good. It is all video based, but it is so much more than a professor lecturing with slides, that it is so busy with graphs and pointers and stuff happening on the screen with quite a fast speaker, that there is no need to speed up the video.

So for me it really boils down to ease of use on the go on a smartphone.  The EdX Berkeley one is out, but it really is a toss-up between the Stanford and Udacity courses.

In the end I picked the Udacity’s Intro to Inferential Statistics as I had already done its Intro to Descriptive statistics. Perhaps some day as a review and to cement in the basics even more, I will come back to the Stanford’s course.

To round-up theoretical material with some lighter material for this sprint I also plan to read the book The Theory That Would Not Die: How Bayes’ Rule Cracked the Enigma Code, Hunted Down Russian Submarines, and Emerged Triumphant from Two Centuries of Controversy by Sharon Bertsch McGrayne and listen to podcast http://www.thetalkingmachines.com/ about Machine Learning. The book has been on my list to read from beginning of year so I will start with that first, but I have received some other great recommendations, so keep them coming!

In my previous post I explained why I am changing the format of my Personal Phd from three month long quarters to six week long sprints.

Today I am starting the second year of the project with its first sprint. Starts today (November 7, 2016) and will last till December 25th, 2016. That is actually seven weeks instead of six weeks due to one week of vacation during Thanksgiving. In the previous year I discovered that I really like to disconnect during vacation and no studying gets done. So this year I want to account for that in the schedule.

The first week Nov 7- Nov 13 is a prep week where I will try out different resources and decide on curriculum.

Five weeks Nov 14-Dec 18 will be 4 weeks of studying and 1 week of vacation.

The final week Dec 19-25 will be a wrap-up week to finish loose ends and write up a summary post.

The main topic for this sprint is Statistics. I started learning Statistics in the third quarter with basics of descriptive statistics, now its time for the basics of inferential statistics.

I recently started listening to podcast The Talking Machines about Machine Learning. So far very interesting, so I plan to listen to all of the episodes from the very beginning for this sprint.

I took a quarter long summer break from my Personal PhD project and I am ready get back into it again!

One thing that I am changing for my second year of Personal PhD is the format.

In the previous year I did 4 quarters each 3 months long, where first 2 weeks were prep, 10 weeks were studying and 2 weeks were wrap-up time. The built-in prep and wrap-up time is super important. Because I get to decide my curriculum I can make the length and rhythm of the periods to be the most optimal for me.

The prep time is for figuring out the curriculum for the next period. It is for trying out few different resources before settling on a specific curriculum.

Even with the best intentions a side project inevitably get delayed so the wrap-up time is a life saver. It’s a time to catch up, finish the last chapters, write up a post summarizing the period. Its a chance to start the next period on time instead of feeling down for delaying it.

From the last year I found that a quarter is a very long time and working on several big things at a time is harder. Also a quarter length project is not agile enough. With shorter periods I can focus on one topic at a time and choose topics as need arises. Also with shorter sprints I can plan better around vacations. From what I saw from last year, I am usually totally disconnected during vacations and get very behind on studying.

So going forward I will do shorter sprints. Each sprint will be six weeks long: one week of prep time (finalizing curriculum and trying out few resources), four weeks for studying, and one week of wrap-up time.

I still plan to write updates, one at the beginning, one in the middle and one in the end.

This update is going to be super short as I need to head out to the airport in 10 minutes for a 10 day vacation trekking in Patagonia and I am not taking my laptop with me. I can’t remember the last time I went somewhere without my laptop. It probably was 6 years ago (for another trekking trip)!

• 45 day streak on Duolingo and Memrise apps to learn Spanish
• Finished up to Lesson 4 (incl.) of Udacity’s Intro to Descriptive Stats Course.
• Continuing on the stats theme, I finished reading book  Superforecasting: The Art and Science of Prediction by Philip E. Tetlock and Dan Gardner
• Watched videos from Weeks 1 and 2 of Coursera’s course on Natural Language Processing https://www.coursera.org/course/nlangp (this wasn’t part of planned curriculum, but a coworker wanted to study it so we decided to join forces).
• Read another chapter (chap 13) from Deep Learning book.

I am having a hard time to dedicate a regular time for studying. Last year somehow my schedule was more consistent and regular. Now it is less so. Seems now my schedule is more project based and focusing on one thing at a time (which could be traveling, or dancing, exercising, or doing big spring cleaning). That’s ok though 🙂 I can do the same with studying for my Personal PhD and do batches of studying every two weeks. Will see how this system works for me.

Now onto the update.

So for this quarter so far I have done 3.5 lessons (out of 7) from Udacity’s Intro to Statistics course.  It is quite basic and at sometimes can feel too slow paced, but I think it will instill really solid basics.

I finished reading The Signal and the Noise: Why So Many Predictions Fail — But Some Don’t  by Nate Silver. It was fascinating read and went into history of IBM Deep Blue winning in chess, history of weather forecasting, history of earthquake forecasting and poker. Also had a very good explanation of applied Bayesian statistics to making and adjusting predictions.

I have restarted studying Spanish. I am using Memrise, Duolingo (and now I am on 25 day streak!) and Coffee Break Spanish Season 4 podcast. All this for my upcoming trip to Patagonia at end of March! Woohoo!

For paper reading: I have read chapters 10-12 of Deep Learning book and few other papers.

Computer Science / Math track

As I was going through the Coursera Machine Learning course and learning about Speech Recognition and reading papers I see that one gap in my knowledge is statistics. For example terms like this: Hidden Markov Models, posterior probabilities, Bayesian statistics, cross entropy, Gaussian, negentropy, doubly stochastic process, Cauchy density. I have heard of simpler things like p-values or confidence intervals, but I wouldn’t know how to produce them if I needed.

Therefore I decided that for the CS track I will focus on statistics and probability for the next quarter. (Beauty of making up my own curriculum and being able to adjust real time to the needs I have at work.)

I looked at few stats courses and I have settled on Udacity’s: Intro to Descriptive Stats (https://www.udacity.com/courses/ud827) & Intro to Inferential Stats (https://www.udacity.com/courses/ud201).

Udacity has a mobile app so I can watch videos or do quizzes while on the bus for example, which does add to the convenience. Also, Udacity recently came out with Machine Learning Engineer Nanodegree which looks very interesting, it includes a very new course on Deep Learning in collaboration with Google. (https://www.udacity.com/course/deep-learning–ud730)

To throw in some fun reading about statistics I plan to

* finish reading The Signal and the Noise: Why So Many Predictions Fail — But Some Don’t  by Nate Silver (amazon)

* read Supreforecasting: The Art and Science of Prediction by Tetlock and Gardner (amazon)

I commit to finishing Intro to Descriptive stats course for this quarter and will see how far I will get with the rest.

Linguistics track

Paper Reading

Writing

Computer Science track

Linguistics track

 Writing

Summary

P.S. The formatting of this post is whacky, but I rather get it out imperfect than not do it at all.