Posts

Educatioinal Resources fo teh Ising Model

The short of it:

Details the theory and resources that resulting from an Ising model project developed for a graduate Statistical mechanics course at the University of Pittsburgh. The main assignment is provided by a Jupyter notebook in which a framework to the Monte Carlo Modeling of the 2-D Ising Model is given and students are walked through developing the missing pieces. A complimentary website has smaller sections which get students accustomed to Python, installing necessary software, and working with Jupyter notebooks.

Background

1. Ising Model for education The Ising Model is a useful tool to teach many important statistical mechanics concepts

   • Phase Transitions! Here we’re able to see a symmetry breaking phase at low temperatures! Additionally, the 2-D Ising model shows a critical point.
   • Although it is a rather simple model, it let’s us understand the behavior of many more systems.
   • Monte Carlo Methods, this allows for an understanding of configuration space, statistics, a sampling scheme used to solve many real world problems. Thus in addition to just understanding Monte Carlo methods, here a framework Python class is built in which students will help build in a few pieces of code, such as a Metropolis-Hastings algorithm or energy evaluations.

2. Ising model theory TBA

3. Our assignment format

The main assignment is provided by a Jupyter notebook in which a framework to the Monte Carlo Modeling of the 2-D Ising Model is given and students are walked through developing the missing pieces. The Jupyter notebook is hosted on github here. A complimentary website has smaller sections which get students accustomed to Python, installing necessary software, and working with Jupyter notebooks.

The way in which this was easiest to send to classes was through google drive. Each student was provided with a link to their own folder containing the notebook, precomputed results for some of the longer simulations, and a results folder where any plots generated would live. After students had completed the notebook, a Google form was used to evaluate their understanding of the 2-D Ising Model and it’s connection with broader concepts.

We had fun putting this together, and we’re still working on cleaning eveyrthing up. Even in this 0.1 version, there is still a lot to gain from the materials!

Chemreps

The short of it:

Chemreps is a python package which aims to create molecular representation as presented in the literature that can be used to generate machine learning models.

Background

Machine learning has become a highly viable tool within the quantum chemistry field. The hopes are that one can learn predict the surface of a chemical space while avoiding the need to calculate each and every molecule with a high-scaling method such as couple-cluster singles and doubles with perturbative triples (CCSD(T)). The performance of the machine learning model is dependent on the type of fittin approach used, Kernel ridge regression or SVR for example, and the form of the data. For chemistry, one needs to capture relevant structure information. However, there is no agreed upon singular approach for a descriptor. This has lead to many different descriptors entering the arena.

Barefoot Pythontessa

Although I intended to keep this a purely computational chemistry related blog, sometimes you do some fun things that can be shared. I was feeling a little stressed and decided I wanted to try my hand at cooking, which is usually a disaster. However, the most stressful part was finding a recipe. So instead of living in the stone age and writing the recipes down on notecards that inevitably get ingredients spilled on them and destroyed, I turned to Python. I categorize the recipes into three dictionaries: entrees, breakfast, dessert. From there I store the recipe URL under a second key in the nested dictionary.

The script is basically a glorified bookmarks bar, i.e. it’s very simple and pretty dumb, but also maybe useful to someone.

You can download the code here