About Me

 I am a... well, mainly a physicist. 

What does a physicist do? That's the question isn't it? What do we do? 

In Malaysia, most physicists end up teaching in a University. 90% of their physics research is in Materials Science. The stuff like semiconductors, thin films (which you use in touchscreens), and these days, nanotubes and graphene. 

There are a few of us going in other directions. Some are interested in Radio Astronomy, another in photonics, a bunch go into GIS and Remote Sensing. 

But there aren't that many physicists in total. If 90% go into Materials Science, that leaves just a few dozen interested in other fields. 

What about me? I am, mainly, a physicist. I can code.

I like to model and simulate physical systems. Oscillating springs, solar system animations. I use Vpython for that. In this example, I simulated a problem in Orbital Mechanics. Suppose you want to travel from one planet to another? Your flight will not be in a straight line because of the Sun's gravitational pull. It will follow an elliptical path, which in this case is called the Hohmann Transfer.

https://www.glowscript.org/#/user/zorniy/folder/Public/program/newhohmann

 

 

 

 

 

 

I also am interested in machine learning. How a computer can tell the difference between several images, for example. Below, I show some things I did for fun. The first is a brute-force hand-coded Nearest Neighbor Classifier. Slow, but I'm proud of it. I know there's a library in Python that can implement Nearest Neighbour faster, but let me have my moment. Imagine how proud the caveman was who invented the wheel!

KNN is a form of unsupervised learning used for clustering elements (in this case, clustering pixels of similar colour). The pixels are assigned to the "centroids" closest to them. Then the average of the pixels associated with a centroid is calculated and used to update the centroid, so the centroid shifts a bit, and the process is repeated for several iterations or until the cetroids more or less stop moving.

Support Vector Machine (SVM) is a supervised classifier method, where a "machine" is taught to associate training data, which are subsets of a main data set, with certain targets. This trained machine can then be used to classify the main data set. 

Imagine you have two flocks of sheep, black and white. You want to build a fence that is an optimal straight line between them. So you pick out some black and some white sheep and use their positions to determine the best fence position and orientation. When you build the fence, hopefully it will separate the two flocks completely! 

What if you have three flocks, A, B and C? Then you need to determine the fences of A vs (B,C), B vs (A,C), and C vs (A,B).

SVM using Python scikit-learn or sklearn This is a program to illustrate how Python can be used for machine learning. sklearn is a Python library which can implement machine learning. There are others such as the more famous tensorflow and keras and PyTorch. I found Jupyter Notebook and Jupyter Lab to "excellente" for this kind of project.

I have also performed wavelet analysis of signals, optimizing the wavelet denoising parameters using the signal's noise statistics. A lot of computer-heavy stuff.

So... am I a programmer, or "merely" a physicist who can code? Am I human, or am I dancer?