another golden quote of the conversation series.

In the labyrinth of academia, where jargon is the king and complexity its crown jewel, I’ve often found myself knee-deep in the weeds of technical details. So when I am in regular meetings with my advisor to talk about papers, it’s often the case that I got trapped into trying to explain all the maths. It’s a dance of numbers and theories, but the elegance of complexity often overshadows the audience - that my advisor often has to interrupt me and shift the topic back to the general background and insight of the researches. ...

As I walked into the grandiose setting of our school’s annual board meeting, a thought struck me - meetings are a peculiar trade-off. We exchange our precious time, hoping to gain something intangible yet potentially valuable. This meeting, with its ceremonial grandeur, was a perfect example to reflect on this paradox. Our school had transformed the largest meeting room into a spectacle. A massive circular table dominated the center, surrounded by a hundred student volunteers, including myself. It resembled less of a meeting and more of a performance, a ceremonial act where real discussions were replaced by pre-decided scripts and roles. ...

Exploring the New Frontier of ChatGPT: A Journey with Specialized GPTs Recently, ChatGPT has undergone a remarkable transformation, introducing a diverse array of GPT models beyond the familiar GPT3.5, GPT4, and DALL-E. These specialized GPTs, likely fine-tuned versions of GPT4, offer tailored functionalities for specific tasks. A Test Drive with the “Data Analysis” Chatbox I ventured into the world of these new models, starting with the “Data Analysis” chatbox. However, my experience was mixed. The processing speed was sluggish, and the output generation seemed unstable. Especially during lengthy data analysis tasks, the system struggled, indicating that it might not yet be fully optimized for complex, long-sequence outputs. Moreover, encountering an error—like a connection loss—meant restarting the entire process from scratch, a time-consuming and frustrating ordeal. ...

The author of this blog has turned 21 years old and can legally drink alcohol from now on! Here are some memories of her brilliant 20th year of her life, as she wanted to share a bit the joy with u as well! November 11, 2022. At my auntie’s, MinHang district, Shanghai. Stayed two weeks in my auntie’s lovely house in MingHang. She cooks excellent crabs and fried pork. It was such a beautiful tranquilent period for a cure. ...

Aggarwal (2019)’s WINE paper proposed a LP-based framework for modelling autobidder for value maximizers. But it does not capture value maximizing bidders with budget and ROI constraints.

This exercise is designed to be accessible yet stimulating, introducing students to the process of bounding the Price of Anarchy (POA).

Should AI-Generated Images Be Copyrightable? Yes and No.

I often encounter a recurring issue when running projects on a remote server and trying to collect data afterward. The setup in my run_script.sh usually includes a line for sending emails once the job is done. #!/bin/bash #BSUB -J StockTraderTestRun # Set the job name to StockTraderTestRun #BSUB -q gpu # Submit to the long queue #BSUB -o /nfsshare/home/tang/StockTrader/%J.out # Standard output file #BSUB -e /nfsshare/home/tang/StockTrader/%J.err # Standard error file #BSUB -N # Send email at job finish #BSUB -u ariana_tang@outlook.com # Set your email address # Change directory to where the main.py script is located # Execute the main.py script cd /nfsshare/home/tang/StockTrader/StockTradingDRL_upload/ python /nfsshare/home/tang/StockTrader/StockTradingDRL_upload/main.py novel_lstm_positive But here’s the rub: the email dispatch never functions as intended. There’s a subtle irony in how my efforts to debug are commensurate with the perceived benefits of receiving an email notification post-job completion. Admittedly, I’ve been somewhat negligent, routinely plugging in the line #BSUB -u ariana_tang@outlook.com # Set your email address and occasionally altering the address, half-hoping for a spontaneous fix. ...

In the thrilling contest of high-frequency trading, my squad is called ‘MONOPOLY’ inspired by Ariana Grande’s hit single of the same name. The song is about “friendship, freedom, protecting your energy and staying right in your bag. We’re all about that life in the lab — working hard and spilling the tea on all things academic. I spent the whole day diving into developing a DRL algorithm for our HFT contest — it’s both a blast and a bit of a time sink, considering opportunity cost. But anyway, I got our code up and running and even churned out some early results. ...