Post #1 5/23/24
The first step in my project is getting a more detailed understanding of the research proposal written by my advisor. In addition, I have begun a cosmology course on MIT courseware to ramp myself up on cosmology which I have not been introduced to before. Finally, my advisor also added me to a group meeting every Thursday and today was the first meeting. Here, I was introduced to postdocs and researchers as they explained their cosmological discoveries through the medium of the papers they had already written.
Post #2 5/30/24
I met with my advisor on Monday (in person) and was allowed to ask any questions that I may have about the physics and the proposal itself. After some deliberation, we decided it was best for me to spend time with the codebase to produce some plot variations to identify important parameters. I was also recommended to read an introduction to cosmology textbook.
Post #3 6/6/24
I was able to produce some plots with variations that my advisor wanted. However, I am to keep on this path to include more variations and more parameters. The goal that was decided was to emulate a plot of 500 simulations found in a paper closely related to the work we are doing.
Post #4 6/13/24
I have succeeded in creating more variation with 3 parameters. However, it is clear that more parameters are necessary to produce the plot of the 500 simulations mentioned above. The goal now is to increase the parameter space to 5 parameters. Much of the trouble has come with a lack of computational power to run many different simulations. Therefore, my parameter guesses require more insight which I have been getting from closely related literature. In conjunction with this project, my advisor asked me to document my use of ChatGPT to better understand its use in this kind of research.
Post #5 6/20/24
This week, in addition to producing more variations in the global 21cm signal, I have been tasked to create a document that summarizes the meaning of a new total of 7 parameters. I found these parameters from papers attempting to emulate the EDGES signal using machine learning.
Post #6 6/27/24
This week, I met with my advisor discuss to the physical meaning of each parameter after doing my own research. Most importantly, the primary goal is to understand how these parameters depend on each other and we hope that the physical meaning of the parameters will shed some light on this. I gained some progress from the insights that my advisor gave to me. My next steps are to describe the qualitative effects of the changing each parameter on the Global 21cm Signal.
Post #7 7/4/24
At this point, I have compiled a document with physical definitions and qualitative effects of the changing the parameters on the signal. All of this was for the aim of our original goal, to simulate the Global 21cm signal using a complete parameter space. With the insight from the document, I plan to write a program that iterates through our predicted parameter space.
Post #8 7/11/24
My advisor was unable to meet this week, however, my work from last week was still in progress. In addition, I was assigned some more papers to read to help me. So, this gave me more time to work on isolating the effects of each parameter, to more methodically and accurately simulate the EDGES signal. I also revisited the physical meaning of each of the parameters in order to define ranges that ensured that my parameter space search would not yield an unphysical result.
Post #9 7/18/24
I was finally able to come visually close to the EDGES signal. At this point, my advisor and I decided that the next step was to create a metric in order to determine how close our simulation was to the EDGES signal, which was the root mean squared difference between the simulation and the EDGES signal. This leads us to a way to search the parameter space with a metric to establish the accuracy of each parameter set.
Post #10 7/25/24
Finally, I spent my time summarizing the work that I have done in a "Jupyter Notebook" to make searching the parameter space very straightforward. The next step is to build a machine learning model that uses my metric and parameter ranges to train, in an effort to methodically search for a possible set of parameters that explain the EDGES signal. While I will not have time to continue this project until completion, my work was a necessary step to begin training an emulator of the signal.
Comments