HNHacker News
TopNewBestAskShowJobs

mailshanx

214 karma · joined March 25, 2012

ML engineer living in Singapore. orthogonal dot eigenvector at google's mail service.

https://sg.linkedin.com/in/mailshanx

submissionscomments
mailshanx··on Thinking the unthinkable
Thats intriguing. It seems like he has somewhat adjusted over time. Did you witness a general improvement in his personality as he grew up? Did things especially improve in his late teens perhaps?
mailshanx··on Ask HN: Are you alone in San Francisco on Thanksgiving?
An awesome gesture. Happy thanksgiving from Singapore!:)
mailshanx··on Early detection of Twitter trends explained
Thanks for the excellent explanation, and many congratulations on your thesis!:)

Could you point to any resources on time series analysis? While i am well familiar with supervised/unsupervised learning methods for tasks like classification, anomaly detection etc, analyzing time series is a different beast. And most machine learning literature (eosl?) doesn't seem to address time series data either.

mailshanx··on Ask HN: Help, HN'er in trouble, we need a laywer in Delhi immediately
Hey Daniel i've emailed you a lawyer contact.
mailshanx··on YC W13 application guide - what is "impressive achievement"
Hey PG, would you regard climbing a ~20,000 ft Himalayan mountain (~single digit known previous attempts) as an impressive achievement?
mailshanx··on Genetic Algorithms in CoffeeScript
I would reserve GAs (at least in their naive form) as a last-resort option. They are extremely inefficient solutions because it doesn't exploit any structure in the problem.

Essentially, a GA solution consists of:

1. Select a random point(s) in the search space, hoping that you have arrived at a sufficiently good solution.

2. If none of the solutions are not good enough, generate a new set of points by combining the best available points and adding a bit of random error.

3. Repeat the above until time has run out/u've hit a good enough solution.

However having said that, GAs can be a good option if 1)not much is known about the function you are trying to optimize, or 2)the crossover/mutation functions are designed to reflect some problem structure, or 3)the search space is small enough.

mailshanx··on The Ph.D Grind: A Ph.D Student Memoir
Awesome read, thanks for taking the time to write it :)
mailshanx··on Ask HN: Ramping up math skills?
I find that at a practical level, linear algebra and probability/stochastic processes are the most valuable and heavily used topics in machine learning. For Linear algebra, i'd recommend Gilbert Strang's book+his MIT OCW lectures. Check out Papoulis's text for probability. It is very dense, and packs in lots of insight per page.
mailshanx··on Peter Thiel's Startup Class 16 Notes: Decoding Ourselves
Really? I find it hard to understand why <em>anyone</em> would consider the idea of life extension as taboo. Wouldn't it be safe to assume that almost everyone, irrespective of your religious or other inclinations, desire to live a long and quality life?
mailshanx··on Ask HN: What skills do I need before tackling The Algorithm Design Manual?
Usually, algorithms/data structures are the second course in a CS sequence. If you can code simple programs, you need nothing more than high-school math to be able to work through Skiena's book. Other than that, i would suggest practicing on TopCoder / code chef and the like.
mailshanx··on 30 minutes a day
Can you elaborate?
mailshanx··on What it takes to build great machine learning products
I think this is pretty accurate. Here is an example from my own thesis research: I'm using machine learning to tune an (underwater) communication link, i.e. decide what modulation / error coding algorithms/parameters will yield good data rates in a dynamic channel.

At first i tried using an off-the-self classifier to figure out which parameters will work well. That failed because by the time i had sampled a decent proportion of the possible parameter values, the channel would change (the number of possible combinations is of the order of a few millions).

Turned out that the real problem is not learning the performance of the available parameters, rather it lies in "learning how to learn": i.e. my ML system needs to adaptively search the space, by responding to the history of previous explorations and their outcomes. This kind of exploration would be effective only with an understanding of how the underlying modulation/coding algorithms work and interact with each other.

mailshanx··on Poll: How long have you been programming?
Im 25. I remember my dad mentioning writing software to compute the trajectory of artillery shells when he was a Navy cadet. He later went on to establish some of the early IT institutes the Navy uses to train it's personel (this was some time in the mid 90s).

I would hang out with my dad in his office, and that's when i first saw computers. Later on, i was exposed to Foxpro programs in high school and two semesters of C in freshman year. But i was too busy fooling around with my buddies to take notice. I was studying EE, so din't really get much programming practice except for some Matlab.

It was only toward my third year in college when i sensed that having good programming skills might be useful. That's when i started doing some project Euler problems. My final year project involved an embedded networked application, and that's when got to know about GCC, linux and the open source world.

I am now a programmer in a research lab, and my masters thesis involves building machine learning models for an underwater acoustic communication system.

mailshanx··on Poll: How long have you been programming?
Holy cow, after reading some of the comments, i just realized that I too am a second generation programmer! My dad wrote some software to compute artillery trajectories while he was a cadet in the Navy. He later went on to establish some of the early IT schools for training navy personel.
mailshanx··on The Best Textbooks on Every Subject
Digital communications by Bernard Sklar. A close second is Wayne Tomasi's book. Both books provide great intuition into the matter with just about the right detail.

Once you have both of those down (and have the requisite linear algebra / probability background), read the texts by Proakis and Gallager. They are very math heavy, (but also very precise and thorough) and though often recommended by university professors, definitely not a good introduction to the subject.

recommendation on the basis of grad-level studies and research on communication / signal processing.

← PreviousPage 6 of 6