Jeongmin Lee

Research scientist at @MetaAI Applied Research.

[Jeongmin Lee @vxdp]

@minjang_kim 네? 이런 도시괴담이 한국 메이져 신문에서 기사로 나왔다구요? 🤯

[Jeongmin Lee @vxdp]

Feel like going back to an exciting phd class! can't wait.

[Kevin Patrick Murphy @sirbayes]

For those of you interested in state-space models, @canyon289 is starting (on 1/21) a reading group on this topic based on my book chapter, and will also dive into our dynamax JAX library. Sign up at https://t.co/9xrHHvUS8I

[Jeongmin Lee @vxdp]

Found the best coffee so far in 2022 at Pittsburgh. It used to be my go to coffee place during my phd, but indeed it has very excellent quality compared to many coffee places in Bay Area!

[Jeongmin Lee @vxdp]

Latte of the day.

[Jeongmin Lee @vxdp]

Positional Encodin…

[Jeongmin Lee @vxdp]

Glad to share that our work "Transformers4Rec: Bridging the Gap between NLP and Sequential / Session-Based Recommendation" is accepted at #RecSys2021!

The work was started during my internship at NVIDIA and it has been great to work with awesome colleagues in Merlin team!

[Gabriel Moreira @gspmoreira]

Very excited to have the paper "Transformers4Rec: Bridging the Gap between NLP and Sequential / Session-Based Recommendation" by @NVIDIA Merlin team accepted at @ACMRecSys! This year acceptance rate was just 18%.
Thanks to my co-authors @Even_Oldridge @SaraRabhi5 @ak_ronay @vxdp

[AerIn @aerinykim]

My team spoke very highly about this blog (and they're also wondering if self-supervised learning could eliminate the need for labeling entirely) so I gave it a read. It was a very well-written, thorough overview of SSL. What stands out the most was it was written by Dr. LeCun,

[Jeongmin Lee @vxdp]

Our new paper accepted at AI in Medicine (AIME) 2021 conf is online https://t.co/zXjrPvvNZQ What's new? When you have event sequence data with high variability like ones form EHRs, do fine-tuning with past history of its own on-the-fly. It will greatly increase predictability.

[Jeongmin Lee @vxdp]

My favorite topic is “Debugging Neural Nets”. It contains many practical tips you will need to unblock yourself when a NN model doesn’t perform. https://t.co/KhhKvdovgE

[Jeongmin Lee @vxdp]

One of the greatest courses that covers both high-level overviews and low-level details on lots of important topics in neural nets & nlp! I took it 2017 and really liked it, especially the quiz. It was a big push to read all the key papers of each topic.

[Graham Neubig @gneubig]

2021 version of CMU "Neural Networks for NLP" slides (https://t.co/X2rd0eHNiW) and videos (https://t.co/6Yi2cZmOLE) are being posted in real time! Check it out for a comprehensive graduate-level class on NLP! New this year: assignment on implementing parts of your own NN toolkit.

[Jeongmin Lee @vxdp]

I highly recommend this team, where I interned last summer: awesome colleagues + super interesting problems + tremendous growth opportunities!

[Even Oldridge @Even_Oldridge]

Want to come work with @benfrederickson, @karlhigley and me building open source tools for recommender systems? We're looking for someone with a good knowledge of the @TensorFlow TFX ecosystem to help us integrate our tools and accelerate it on GPU. https://t.co/dlHmNQIC0v

[Jeongmin Lee @vxdp]

@kchonyc @joinClubhouse It seems so. @joinClubhouse started to be unstable since @kchonyc created the room. After he finished the room, entire clubhouse seems totally down. 😔

[Jeongmin Lee @vxdp]

https://t.co/h9Ovqnx1Cd

[Nils Höhing @nilshoehing]

Loved this Multimodal Machine Learning Course by Louis-Philippe Morency and and @pliang279 at @LTIatCMU! https://t.co/tlQmDHmgRR Great to have such resources available online.

[Jeongmin Lee @vxdp]

Cool work on predicting ICU-patients' hypotension events.

[Joo Heung Yoon @penn1600]

In the midst of COVID... we present a new yet classical machine learning method to predict upcoming hypotension for ICU patients - at @Crit_Care. Differences? 1) More data-driven, 2) We tried hard to make it implementable. https://t.co/kOYvwPY6o1

[Pedro Domingos @pmddomingos]

Here's a 40-minute summary of 15 years of research in Markov logic: https://t.co/7LIStvG9g1

[Jeongmin Lee @vxdp]

@minjang_kim https://t.co/hVHeXFAGzu 맥 키보드에서 바로 외부모니터 밝기조정 + HDMI 음량조정 되는 이 앱 추천드려요.

[Jeongmin Lee @vxdp]

After long searching and debugging, I figured out that I need to have z=x+y.detach(). Setting y.requires_grad=False was not enough. Meanwhile, I found this article very informative to understand the error: https://t.co/XDBMyfNla0

[Jeongmin Lee @vxdp]

Friday night reading: is_leaf under torch.autograd. https://t.co/T1iZjslDao

[Jeongmin Lee @vxdp]

indeed, it was to fix this error: "RuntimeError: leaf variable has been moved into the graph interior" where I have z=x+y, where x is from a model that I want to update its params and y is from another **fixed** model.

[Jeongmin Lee @vxdp]

Friday night reading: is_leaf under torch.autograd. https://t.co/T1iZjslDao

[Jeongmin Lee @vxdp]

Friday night reading: is_leaf under torch.autograd. https://t.co/T1iZjslDao