Theoretical

So far, in my postdoc position, I have published one paper (Climate change and chill accumulation: implications for tree fruit production in cold-winter regions), submitted another (Codling moth pest pressures and pest control efficacy under climate change), and preparing a few others. Two of the papers that are in final stages of preparation are about remote sensing; using satellites to see what farmers are up to!

My tutorial paper got accepted in International Journal of Modern Physics C and, I just wrote a survey on opinion dynamics (https://arxiv.org/abs/2004.05286).

My recent paper is a tutorial of opinion dynamics models. It is on arXiv and ResearchGate as well. Opinion dynamics is about how interacting agents—whether it be humans or birds or computers— evolve by interacting and learning.

#opinionDynamics #socialinteractions #opinionformation #Socialnetworks

Mahalanobis distance explained in three steps, please take a look at my note. And here I also explained why one should do standardization before applying PCA to the data.

I wrote a little note about t-test, p-value and hypothesis testing. p-value is used to test how good or relevant are the regression coefficients among other things. Please take a look.

I just watched ``How to Win a Data Science Competition: Learn from Top Kagglers'' course on Coursera. It is a very exciting, informative, practical course. Topics that are never covered in a classroom setting! Most probably watching it would be motivating and helpful.

The reason I have posted the PDF version of the first three weeks below is that I think going through these would be faster than watching the course videos. Also some people learn better by reading!

Week I, Week II-I, Week II-II, Week III, Week IV. (If the forth week material is heavy and does not appear on the screen, download it!)

A short note on Decision Trees, Bagging, Boosting, Random Forest and Extra-Trees. I have quoted parts of "The Elements of Statistical Learning"  which is a very good book and availably for free, and the "Extremely Randomized Tree" paper. I have tried to collect main points and motivations and differences of the mentioned techniques in just a few pages. Hope you find it useful.

These Notes are from SQL for Data Science by University of California, Davis in Coursera. Hope you find it useful.

Week 1, Week 2, Week 3, Week 4.

Counterfeit Coin Puzzle

This is a famous classical problem which I found recently and it is fun. You are given a two pan fair balance.

Different Versions of the Problem:

1. You are given N identically looking coins of which one is fake. The weight of the fake coin is slightly either lighter or heavier than others.

   • What is the minimum number of weighings you can identify the fake coin?

   • What is the minimum number of weighings you can identify the fake coin and its weight (lighter/heaver)?

2. You are allowed to use the balance K times. What is the maximum number of coins for which you can

   • find the fake coin

   • find the fake coin and its weight.

For each case above, there are two possible ways of weighing. Sequential and non-sequential. In sequential version you can use the result of each step to plan your next move(s). In non-sequential, you must have a plan before starting the weighings and you cannot change your plan half way through.

P.S. Whenever I have time I will solve and write a Python code, and put it on Github, for non-sequential version of the game where K is given and the weight of the fake coin is to be determined.                                                                                                                       Famous versions are sequential where you have 9 or 12 coins. Give it a try before Googling!

On the right you can see a sequential case solution!

Opinion Game

Questions and a New Model

Motivating Questions: How do opinions change? How do social networks shape opinion? Is there a connection between the ground being lost by moderates and the growing gap between the extremes in politics? Has the way in which the Internet mediates influence fundamentally changed the way in which opinions evolve?

Simplified quantitative models of the formation and evolution of opinions on a topic (or several topics) are referred to as opinion games and have been an interest to scientists for at least 7 decades. We have created a new, flexible opinion game model. The added complexity in our family of models has already generated some interesting preliminary results. (To see a short tutorial of our model, please click here.)

Code and Paper

We have developed code anyone can download and play with.                                                                                                                    This is the accepted author manuscript of the paper, An energy-based interaction model for population opinion dynamics with topic coupling. The journal version is accessible on International Journal of Modern Physics C.