The October Review
School is officially back! It’s been just about a year since I was in school, but hopefully, these next two years will be the last of my classes in general because school has been HARD as FUCK! I recall my undergrad years when math used to be equally challenging but fun, but lately, my love for mathematics has been put to the test by the density and difficulty that come with pursuing these quantitative degrees.However, life in Los Angeles has been treating me well in other respects.
I’ve settled into a comfortable routine here in LA. I’ve joined the intramural volleyball team, which has been a blast. Playing trivia with my friends every Tuesday is a fun mental workout on testing my non-existent general knowledge, and the happy hours with my fellow Ph.D. students every other week have helped create a sense of camaraderie. Returning to the student lifestyle feels surreal at times, but I believe this academic environment is where I currently belong. This degree isn’t just a stepping stone for my career; it’s another mountain to climb. While it may be one of the most challenging mountains I’ve faced, I welcome the difficulties, as they push me to excel and are the requisite steps required of me to get to the life I desire. I use wanting to move to New York City often as a source of motivation and that this degree is the only thing standing in my way.
While I have briefly described why I have returned to school, I wanted to highlight in brief those reasons yet again. When i began wokring earlier this year, I sometimes felt stagnant in my learning. Therefore at the expense of being a student I am now in an environment yet again full of academic stress but with the rewarding experience of learning something new every day. School also brings an unpredictabilty factor which makes it pretty exciting in practice like local vendors Lulu Lemon, In-N-Out and many others giving away free things all the time on our campus. This transition back to the world of academics has had its share of challenges in the getting back into the groove of things, but I firmly believe that hardships build character. While I might not think I need more character development, I embrace this journey with unwavering commitment to giving it my all. I treat this whole situation as yet another level up in difficulty where we are in the final stages at the Very Difficult mode.
Beyond the academic realm, I’ve finally achieved a long-standing goal of staying in one place for an extended period. Over the past year, it often felt like I was moving every six months as I traveled around the world. However, this time, I’m here for the long run (nearly five years, and hopefully not a day longer). This stability means I can start building lasting friendships and establish my adult life in the vibrant city of Los Angeles yet again. In just my first month back, I’ve already met so many wonderful people, and I can’t wait to share some of the highlights from this month alone. Being in Los Angeles is also exciting in the sense that it is one of the most urban cities in the world. There is nothing short of things to do in this place and I know that I will get many experiences out here because like school, there are many unpredictable factors by living in this city alone. Growing as an adult while I tackle my academics is equally important to me ensuring I do my absolute best in both these aspects of my life.
While school and academic pursuits have been demanding this past month, my overall well-being has been great. I’m in a positive headspace, and I hope to maintain this throughout my Ph.D. journey. So, without further ado, let’s delve into the highlights of this month.
Top Three Highlights of the Month
1. Collette
This month, I would like to start by highlighting one of my favorite parts of the month: meeting Collette Lee. Collette and I began going out towards the end of September and have continued to spend time together throughout October. In the short time I have known her, I have experienced nothing but positivity and joy when she’s around. As my readers, you have implicitly followed my life over the past year, including my dynamic dating experiences. However, truth be told, this dynamic dating lifestyle is drawing to a close, not because I can’t meet people naturally, but because I genuinely feel great about my relationship with Collette.
Collette and I initially met on Hinge, which I believe reflects more on our generation’s dating preferences than any inability on my part to meet people in person. Collette is a breath of fresh air. Spending time with her has kept me on my toes in all aspects of life, and it’s been a while since I’ve gone out with someone who makes my heart race with nervous excitement. Despite my natural confidence, I feel extra nervous, which I believe speaks to the remarkable person Collette is. She is a UCLA alumna currently working in the Design industry, and combining our two worlds has been an incredibly enjoyable experience. In a partnership, compatibility is paramount, and I believe we strike a harmonious balance in our dynamics. She tends to be on the quieter side, in contrast to my ambitious and outgoing nature. However, this contrast is perfect for both of us, as she grounds my lofty ideas, while I encourage her to step out of her comfort zone and try new things.
Our lifestyles also align quite well. She, like myself, is in the process of transitioning into adulthood after graduating. This synchronicity in our life stages has been advantageous, as we are both navigating the challenges of adulthood and supporting each other along the way. While we are both borderline workaholics in our respective fields, we also maintain a strong sense of mutual respect and personal boundaries. It’s truly refreshing to be with someone who may not fully understand my work but genuinely listens to and appreciates what I have to say. I equally enjoy listening to her stories about photoshoots and product launches, which feel like a completely different universe from my own experiences. These conversations have underscored the value of being with someone whose professional life is drastically different from mine, demonstrating that the working world offers more diversity than what is often depicted in TV shows (although I teasingly tell her that her life feels like a TV show at times).
In other aspects, I have nothing but great things to say about her appearance and style. As a designer, she has a great sense of style, which has prompted me to take better care of my personal appearance whenever I am around her. So far, it has worked out well because she has commented on liking my personal style, which is nice to hear. I could go on and on about what a month it has been with Collette, but I thought I would share just a memory or two about some things that have happened. But before I dive in, I just have to say that it’s safe to say that I have a massive crush on this girl.
Therefore, through all the memories we shared this month, I would like to share two in particular. The first highlight of the month was one of our earlier dates, which involved making Kimchi fried rice or Kimchi-bokkeum-bap together. I am not one to cook all the time, but together we picked out some ingredients to make Kimchi fried rice. We added pork belly, chives, and more to create a meal that felt like home. One of the additional nice things about Collette is that she is very open to trying most things, which makes these dates really fun in general.
Another highlight from the month was when we spent the day together going to Rancho Palos Verdes near Long Beach. In addition to her amazing self, she also has a car, which makes LA very accessible. I had always wanted to go to Rancho Palos Verdes because from what I had heard, it was a quiet and beautiful beach town. So together, we ventured down south to catch the sunset. It was both very lovely and enjoyable as we got to appreciate the natural beauty of living in Los Angeles. Afterward, we stopped by a place just north of Rancho Palos Verdes to get some Korean Fried Chicken. It was a craving I had for months, and I was happy to share a pleasant meal with her.
All I have to say is that Collette is amazing, and I feel lucky and really, really happy to have gotten to know her and be with her this past month. Life has only gotten better on that front, and I cherish every moment I spent with her.
2. Halloween
Halloween, no doubt my favorite college holiday, has always been the best part of the school year. The excitement it stirs up among students is unmatched, as it’s the perfect reason for a big celebration. This year was no different, even with my intense graduate studies. Determined to enjoy the Halloween spirit, I made sure to take a break from my demanding studies. I had been looking forward to this for months, especially after missing out last year while I was stuck in Switzerland (not literally). Unfortunately, Halloween isn’t widely celebrated there, so the desire to immerse myself in the festivities this year had become intense.
Despite the busy academic schedule, this fall quarter seemed to shine with several joyful occasions, like the upcoming Thanksgiving’s Friendsgiving and the warmth of Christmas with its enjoyable Secret Santa exchanges.
However, the ongoing challenge of finding the perfect costume was on my mind. The pressure to come up with something new every year was real, but luckily, the nearby Spirit Halloween store saved the day. Despite my hectic schedule, I managed to visit it on the Thursday before the Halloween weekend. Faced with the choice of two different outfits, I went for a playful penguin costume for one night and the iconic Naruto for the next, ensuring back-to-back fun.
The first night was a whirlwind of excitement as my Ph.D. friend Arabdah and I went on a Santa Monica bar crawl. Maybe I had a bit too much to drink at his place, as I felt uncomfortably warm in my penguin suit. However, undeterred, we ended up at Barney’s Beanery for some enjoyable karaoke. I vaguely remember singing a few songs, but the night continued as we moved on to the next bar. I don’t recall much from that stop, but the night ended back at UCLA, where I met up with Collette, ending the night on a happy note, despite not getting into the popular Rocco’s Tavern.
On the second night, nursing a stubborn hangover, I wondered if it was wise to go out again. But any thoughts of rest disappeared as I found myself going to the UCLA football game with Arabdah and Tim Lindsey. Deion Sanders and the CU Boulder team, despite their national following, failed to impress, leaving me relatively indifferent to the sport. It was a cool experience overall to be able to see this nationally recognized team but the effort it took to go to this game makes it a one and done situation. After the game, I quickly switched to an evening of pre-gaming with my friends Jieyi, Xinyu, and Jade, before we went to another club in Westwood. Surprisingly, the night went on until 2 am, a testament to how fun and energetic it was, despite my usual early exits and the exhaustion from the day’s activities. Towards the middle of the night, I ran into my JSA friends Mai and Airi, and we danced until the early hours, finally giving in to the tiredness of the night.
As the weekend came to an end, I couldn’t help but think about the whirlwind of fun and excitement that captured the essence of Halloween, making it one of my most cherished and memorable holidays yet. Unfortunately for me though, it is back to work and all the pressure that brings…
3. ML4H
The final update of the month involves some of my academic work thus far. Over the summer, I was able to submit my first ever conference proceedings for review and it is a new type of excitement that I am not totally used to. Seeing your projects grow into fruition is a surge of new excitement that is very different than other types of things. Therefore this month was a lot of fine-tuning and crunching numbers for our submitted method to hopefully get the opportunity to present my work. This part of the blog might be more technical than other components but I’ll try to alleviate as much technical knowledge as possible. If accepted, I can look forward to going to New Orleans, Louisiana this upcoming December.
Multimodal Clinical pseudo-notes for EHR prediction
Abstract:
Electronic Health Records (EHR) are comprehensive databases containing multimodal information about a patient’s health history. In recent years, Transformer-based models have shown promise in various downstream tasks, including mortality prediction and diagnosis. However, these approaches have either considered one component of EHR, or considered its multiple components as a single data modality. In this work, we treat EHR as multimodal, separately representing concepts like diagnoses, medications, procedures, and lab values. Our novel ``pseudo-notes” method transforms these modalities into structured language texts, allowing us to leverage general Large Language Models (LLMs) for individual EHR representation from the MIMIC-IV database. We test our model by predicting Emergency Department (ED) Disposition and find our multimodal model outperforms against a single modality method, and other machine learning methods, demonstrating its effectiveness.
Detailed Explanation
Large Language Models have made its presence in our world with the likes of ChatGPT, DALL-E, and other robust and impressive models. Part of its impressive capabilities involves its high accuracy of various predictive tasks. Therefore, our work poses a new view to classification problems (prediciton) for medical-related data.
Traditionally, data is stored in databases called Electronic Health Records (EHR) and it is stored in a tabular data structure (excel sheet like). Within these tabular data structures are medical values, drug codes, and various other numerical and textual data all within this database. Traditional Machine Learning methods like XGBoost and Deep Neural Networks are exceptional at taking tabular data and performing really well on prediction tasks. However, these methods tend to struggle when the dimensionality of our data is really high. EHR falls under this high-dimensional data because computers are unable to process texts when presented in this tabular form. Therefore, some solutions that are typical in machine learning are creating binary labels called one hot encodings where if a patient X has disease A, they would have a 1 in that disease column and a 0 if they didn’t have that disease. However, within a hospital system, there are hundreds of thousands of unique diseases that make this difficult for computers to ingest.
Therefore, our work aligns with the recent advances of our technological world and follows a language-based approach where special algorithms are able to take in text data and make meaningful representations out of them. We no longer are faced with the dimensionality problem because each disease can now be represented in its canonical textual form and can therefore take in multiple sources of information. So in addition to diseases, we take the following EHR concepts (modalities) which will be used for our task: arrival information*, capturing patient demographics and means of arrival; triage, capturing patient vitals and complaints recorded when they first arrived, medication reconciliation (medrecon), the list of prior and current medications the patient was taking; and measurements collected throughout their ED stay including patient vitals; laboratory values; and medications received pyxis.
Our goal of this project is to determine whether a patient will be sent to the hospital or sent home when they show up to the emergency department (ED Disposition).
We utilized a BERT model which is very good at textual data ingestion before feeding it to a transformer model which is able to learn the combined data and predict the correct emergency disposition. We found that our model was able to accurately predict ED disposition with a 94% accuracy which continues to show the synergy between healthcare and artificial intelligence systems. Our hope is to round out this work in the coming months and get it sent to the ICML Conference Journal which is a presitigious machine learning conference. The hope is then to present our almost finalized work and personally get an opportunity to return back to Europe this following July. So while school has been relatively stressful and busy, I am excited about the progression I am making in the research front.
One thing I learned this Month
Graduate School is Actually Hard
I don’t know whether it was the gap year or my knowledge gaps, but pursuing an interdisciplinary graduate degree has proven to be quite challenging in practice. Computational Medicine falls within the realm of biomathematics coursework, and the amalgamation of biology, mathematics, statistics, and computation has presented a rather rough start to the school year. Similar to my experiences at EPFL in Switzerland, academic life here is, quite simply, demanding.
Coming from a mathematical background, one might assume that statistics and computation are already in my wheelhouse, but this degree involves so many diverse components that make it exceptionally challenging. For instance, the nuances between mathematics and applied mathematics, while not overtly apparent, cover significantly different aspects. In the coursework I’m undertaking this quarter, they’ve opted for a more pure mathematics approach, which has, in turn, given me quite a headache in all my subjects. What compounds the difficulty is that everyone in my cohort comes from varying backgrounds, so none of us find this material to be a walk in the park. Allow me to briefly introduce my cohort members and their backgrounds to provide context for the challenges we’re grappling with:
- Simon Lee (Me): Applied Mathematics/Computer Science
- Zach Schlamowitz: Pure Mathematics
- Sayun Mao: Biophysics
- Arabdah Biswas : Biomathematics
- Yihui Cen: Informatics and Biology
- Hank: Biostatistics
It’s almost as if the department intentionally assembled our group for these reasons. Each of us has strengths and weaknesses that complement one another. At times, though, I feel like I contribute the least, especially in understanding the theory and derivations – areas where I struggle while excelling in practical examples. Unfortunately, for our program, theory and derivations are crucial aspects of our coursework. This became especially evident when I enrolled in an algorithms course, which I expected to be similar to other computer science courses I had taken during my undergraduate studies. However, to my dismay, it leaned more towards a pure mathematics proofs course, where we prove A) why the algorithm works and B) why it’s a good algorithm. I’m not entirely certain of the practicality of these two objectives, but it’s something I’m working to adapt to because we’re all quite apprehensive about the first-year comprehensive written exam. Our Computational Medicine Department has structured our Ph.D. program to mimic an Applied Mathematics Ph.D. program, which includes a daunting written exam that we all must face in the middle of the summer.
While there have been early concerns this month regarding the difficulty of graduate school, I’ve realized that it might be the most significant obstacle on my path to obtaining my degree. Beyond the demanding coursework, there have been some initial research successes in my work on a rather hot area of research involving LLMs and NLP. So, throughout the year, I suppose I can continue to provide updates on my journey, but it’s been somewhat disheartening to see how much I either don’t know or have forgotten since leaving school a year ago.
One thing I’m banking on, however, is my work ethic and my ability to persevere. This quarter, I’m tackling the most mathematically intensive courses I’ve ever taken at UCLA, and I’m embracing the challenges that lie ahead. It’s also worth noting that if these courses were easy, everyone would have a Ph.D., so the difficulty is par for the course. I sincerely hope that the countless hours I’m expecting to invest in these courses will ultimately pay off in the long run. After all, obtaining my Applied Mathematics degree left me somewhat underwhelmed, as I didn’t become the mathematical wizard I had hoped to be. However, a Ph.D. program, with its rigor, might just provide me with the opportunity to achieve that feat.
Outside of my studies, the first month has been a bit of fun. Balancing social and academic life has once again proven to be a challenge, but as someone who thrives with both, I’ve found a way to make it work. Furthermore, from a more general perspective, attending classes as a graduate student feels substantially different from my time as an undergraduate. The class sizes are much smaller, and they really encourage you to delve deeply into the intricacies of each subject. While I don’t feel like Robert J. Oppenheimer or Alan Turing in their prime, it’s been quite an experience to pursue a Ph.D., arguably the most challenging of all degrees. I just need to remember to breathe and take care of myself because, sadly, poor dietary choices have crept in again due to time constraints.
Quote of the Monthly
“Patience is a Virtue”
Contact
Simon Lee
simonlee711@gmail.com
simonlee711@g.ucla.edu