Research (Formal and Informal)

a collection of mini-papers written by Angela Yang

  • Forward-Backward is just Backprop


    ~15 minutes

    This is the final project paper for the Computational Linguistics II course at UCLA. It is based heavily on the paper "Inside-Outside and Forward-Backward Algorithms Are Just Backprop (Tutorial Paper)" By Jason Eisner. This paper goes through the derivation of the Forward-Backward algorithm in the context of HMMs and PCFGs, and shows that it can be generalized to backpropagation. Click Here to View In Browser

  • Natural Language is Not Context-Free: A Comparison Between Linear-Indexed Grammars (LIGs) and Tree-Adjoining Grammars (TAGs)


    ~15 minutes

    This is the final project for the Computational Linguistics I class, on the mildly context sensitive nature of natural language. This paper goes through examples and limitations of Linear-Indexed Grammars and Tree-Adjoining Grammars. Click Here to View In Browser

  • Tone Sandhi in Taiwanese Hokkien


    ~15 minutes

    I wrote this term paper as the final project for my Phonology I class, on the phonemes of Ukrainian. I consulted a native speaker of the Taiwanese Southern Min Dialect, Jason Yang for this paper. The analysis is based on evidence gathered from recordings of the consultant. Click Here to View In Browser

  • Phonetic Study on Ukrainian


    ~15 minutes

    I wrote this term paper as the final project for my Introduction to General Phonetics class, on the phonemes of Ukrainian. I consulted a native speaker of the western Ukrainian dialect, Maksym Prokopovych for this paper. The analysis is based on evidence gathered from recordings of the consultant. Click Here to View In Browser

  • Comparison Analysis on Rust and C Through the Lens of GNU grep Implementation


    ~5 minutes

    This is the final paper for my Programming Languages class, on the improvements that Rust has attempted to make on C++ through the topic of grep. Click Here to View In Browser

  • The Effects of Citric Acid on Contraction of Mouse Cardiomyocytes

    this paper was written 8/12/2019, reflection was written 11/15/2022

    ~5 minutes

    Link to the paper
    Reflection: I wrote this experiment in tenth grade of high school while interning at Dr. Vondriska's Heart Failure Lab. I proposed a solution in which adding Citric Acid by way of perfusing through the coronary arteries of a mouse heart, would stimulate contraction of the tissue. While my experiment was a partial success, supporting my hypothesis, with further research and a few biology classes, I now know my hypothesis was naive and flawed. Here is my reflection after three years:

    1. The Energy Problem:
    Context and Approaches: The main issue with the sustained contraction of heart cells is that like most cell activities, contraction of the actin and myosin filaments in a cardiomyocyte required energy in the form of ATP or a similar nucleotide triphosphate. The issue was that being a short-term energy storage molecule, ATP tends to be unstable, and thus difficult to synthesize and costly to purchase. Direct addition of ATP salts was off the table for this reason. An alternative solution was offering glucose directly to the cell buffer for cells to uptake. However, this also runs into the issue of glucose uptake requiring insulin to cross cell membranes. This also why diabetes treatment required more than just sugar regulation. Insulin was less costly than ATP salts but also beyond my project budget and timeline. This is where my final (and flawed) solution came in. I sought to hijack the ATP production pipeline by providing an intermediate in the buffer so the cell could pick up from the middle of the cellular respiration process and create its own ATP without requiring insulin.

    What was wrong with it?: Firstly, just as glucose—a large molecule by the standards of the lipid bilayer—requires a insulin ligand-gated transporter, many other molecules used by the cellular respiration pipeline likewise need special transporters. I should've researched this more thoroughly before choosing my intermediate at the very least. Moreover, glucose only requires transport across the cell membrane as it begins the cycle of glycolysis which occurs in the cytoplasm. It is immediately converted to a geometric isomer, fructose, which is unable to cross the cell membrane completely. This maintains a favorable passive glucose gradient, and prevents loss of already acquired glucose. In order for glycolysis to function, other intermediates before the next stage of cellular respiration must also be membrane impermeable. This changes when we reach the Citric Acid Cycle, which occurs within the mitochondria. These intermediates now are unable to cross the mitochondrial membranes as this would result in a huge loss and inefficiency in the cellular respiration process. My issue was actually twofold: figuring out how to hijack the cellular respiration system, but also figuring out how to get across respective membranes in order to allow cells to utilize my chosen substrates.

    2. The False Positive:
    As stated, this experiment was a partial success, as my hypothesis was supported by my results. However, for the reasons above, this is an impossibility. Instead, what I had was a false positive hypothesis. In fact, the null hypothesis of this experiment turned out to be the one better supported. The null hypothesis is the hypothesis that there is no correlation between the dependent and independent variables. I ran four experiments: whole mouse heart with my citric acid buffer, one without, isolated cardiomyocytes with buffer, and another without. With the whole mouse heart, both samples continued contracting with citric acid and without. This supported the null hypothesis. However, with isolated myocytes, they only contracted with my buffer. This is what I assumed supported my hypothesis, but this is an assumption in the face of a clear falsity with whole heart experiments.

    3. Technical Flaws:
    As with all experiments done in physical wet labs, it is prone to researcher or methodology error. Let's target the one likely responsible for the false positive result I got. The isolated cardiomyocytes were beating with my buffer but not visibly so without it, whereas both whole hearts beat the same. There could be several reasons for this:

    a. Small Sample Size: While I did observe contractions, I only observed a small proportion of cells present contracting. Most weren't contracting in both samples.

    b. Centrifugal Damage: In order to isolate the cardiomyocytes from the rest of the mouse heart tissue, I ran the tissue through the centrifuge with a cell-separating buffer. This was quite destructive, observably so, as many of those noncontractile cells were shriveled from the physical force. This could've also caused unforseen results such as damaging the cell membrane to increase permeability and allow the buffer to actually take effect when it normally wouldn't have.

    c. Incomplete Perfusing: On the flip side, the whole heart could've not expressed the hypothetical results as the perfusal of the heart by the Langendorf apparatus may not have been sufficient as pressure in the smaller coronaries could've meant buffer was more likely to flow out of the aorta on the outside of the heart rather than through its vasculature.

  • copyright 2022 Angela Yang