Ji Won Park @ICML2024
@jiwoncpark
Followers
779
Following
1,031
Media
29
Statuses
495
Principal ML Scientist at @Genentech @PrescientDesign . Bayesian optimization, Bayesian inference.
Foster City, CA
Joined January 2014
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
jaehyun
• 466269 Tweets
LINGORM TSOU FINAL EP
• 311404 Tweets
#ใจซ่อนรักตอนจบ
• 289266 Tweets
SEVENTEEN
• 252447 Tweets
jeonghan
• 230651 Tweets
LISA X ELLE MAG COVER
• 95080 Tweets
#海のはじまり
• 61603 Tweets
#JAEMINBornToBeLoved
• 57388 Tweets
Abdullah
• 56856 Tweets
#사랑하는_재민이의_스물넷을_찰칵
• 55629 Tweets
最高の夏
• 54803 Tweets
ペルセウス座流星群
• 53304 Tweets
BOTO PARA SA SB19
• 50872 Tweets
Leicester Square
• 49728 Tweets
RadhaSoami Vs Sant Rampal Ji
• 47557 Tweets
CDTV
• 38875 Tweets
song for you
• 35174 Tweets
junhui
• 29421 Tweets
オーロラ
• 27397 Tweets
HAPPY BIRTHDAY JAEMIN
• 24768 Tweets
PHUWIN SUMMERNIGHT EP2
• 24660 Tweets
梨海高校
• 22512 Tweets
AI-5
• 21848 Tweets
DUNK SMN EP2
• 18851 Tweets
CRAZY ALBUM TRAILER
• 17068 Tweets
サカナクション
• 16079 Tweets
Leao
• 14650 Tweets
Wahi
• 13968 Tweets
海ちゃん
• 13384 Tweets
津野くん
• 12595 Tweets
ジョンハン
• 10658 Tweets
We are hiring! Apply to our summer ML internship at the beautiful Genentech campus in South San Francisco: 🧵 1/5
7
34
163
Paper alert🚨 Proteins are my new dark energy, but my latest paper employs hierarchical Bayesian inference, importance sampling, deep sets – all methods close to my heart! 🧵
4
18
98
Paper alert🚨 We apply latent stochastic differential equations (SDEs) to model the time variability of extreme astronomical objects called active galactic nuclei (AGN). Check out our spotlight at
#ICLR2023
#Physics4ML
! 🧵
1
11
96
We are excited to present PropertyDAG: Multi-objective Bayesian optimization of partially ordered, mixed-variable properties for biological sequence design! 🧵
2
27
78
PropertyDAG has been accepted to this year's NeurIPS
@AI_for_Science
!
We are excited to present PropertyDAG: Multi-objective Bayesian optimization of partially ordered, mixed-variable properties for biological sequence design! 🧵
2
27
78
0
4
24
If you're at
#SIAMUQ24
, come to my talk on BOtied 🎀, a new CDF-based acquisition function for multi-objective BayesOpt (feat. copulas, tails) in MS226 (Fri 5-7pm)! joint work with
@tagasovska
@stephenrra
@kchonyc
1
5
19
Yesss come talk to us!
#GHC22
@PrescientDesign
is at the
@genentech
booth in
#GHC22
come stop by to hear about biotech, AI and how to help people!
0
3
8
0
3
16
Shout-out to my wonderful collaborators
@samuel_stanton_
Saeed Saremi
@amw_stanford
@HenriDwyer
@GligorijevicV
@RichBonneauNYU
@stephenrra
@kchonyc
here at
@PrescientDesign
@genentech
@Roche
(Come work with us!)
0
1
9
You will join me,
@m_kirchmeyer
, and the
@PrescientDesign
, Genentech team in modeling dynamical systems with ML for drug discovery. We welcome diverse approaches, including those using causality, new solvers, or LLMs! 2/5
1
0
6
This was a project with my amazing co-first author
@joshua_fagin
! We thank Henry Best,
@saavikford
, Matthew Graham, Ashley Villar,
@cosmo_shirley
, James Chan,
@matt_of_earth
and also
@DavidDuvenaud
@lxuechen
for their helpful suggestions throughout the project.
1
0
6
We release with this paper the open-source
@LSSTDESC
Python package, Node to Joy. It implements the structure-enhanced raytracing, training and evaluation of the BGNN, comparison with number counts matching, and the hierarchical inference analysis.
1
0
7
How to ace the GRE and Physics GRE – Jaan Altosaar
0
1
6
By encapsulating our experimental and biological priors on the relationship between molecular properties, our method promises to accelerate computational drug discovery. ⏩💊
1
1
5
Check out my new paper, [2012.00042] Large-Scale Gravitational Lens Modeling with Bayesian Neural Networks for Accurate and Precise Inference of the Hubble Constant
2
0
5
Being so luminous, AGNs can be observed out to great distances, close to the edge of the observable Universe. By inferring black hole physics from AGN light, we can gain insights about the origin and evolution of the Universe, including the nature of dark energy and dark matter.
2
0
5
Simple way to measure cell execution time in ipython notebook
0
0
4
We run multi-objective BO with PropertyDAG over multiple simulated active learning iterations on a variety of tasks, including an antibody design task based on a real-world dataset of expression and affinity measurements.
1
1
3
We modify the latent SDE (~an infinite-dimensional VAE with an SDE-induced process as the latent) to jointly (1) reconstruct the multivariate AGN time series from limited observations and (2) infer the posterior PDF over key black hole properties.
1
0
4
Special thanks to my collaborators
@sibirrer
,
@madiueland
,
@MilesCranmer
, Adriano Agnello, Sebastian Wagner-Carena,
@drphilmarshall
,
@AaronRoodman
, pub board and internal reviewers
@LSSTDESC
🙏
0
0
4
Bayesian optimization (BO) is a sample-efficient framework for navigating the exploration-exploitation trade-off in the vast design space of biological sequences. PropertyDAG operates on top of the traditional multi-objective BO to confer hierarchy to the objectives.
1
1
3
We recover the true input H0 at 0.7% precision, from our test set of 200 simulated lenses. Our pipeline is implemented in the open-source Python package "H0rton":
1
0
4
Traditionally, the external convergence has been estimated by reducing the whole sightline of data to some summary statistics (such as galaxy number counts N) and matching them to a simulation with known convergence. BGNNs can instead take in all available info. No info wasted!
1
0
4
This is a 3-month full-time internship with a substantial ML research component. We will provide you with the data, computing resources, and mentorship to develop algorithms aimed at solving complex biological problems. 4/5
2
0
3
On realistic, physics-driven simulations of AGN time series, we outperform the Gaussian process baseline in reconstruction and can also precisely infer the black hole properties.
1
0
3
To impose a desired partial ordering on the objectives, e.g. expression --> affinity, our framework modifies the posterior inference procedure within standard BO in two ways.
1
1
2
For instance, when designing antibodies, we'd like to maximize the binding affinity to a target antigen only if expression happens in live cell culture -- modeling the experimental dependency in which affinity can only be measured for antibodies that express in viable quantities.
1
1
3
Often, we want to prioritize some regions of the Pareto frontier over others according to a specified partial ordering in the objectives.
1
1
3
We propose the Bayesian graph neural network (BGNN), a deep set that estimates the posterior on convergence by updating local embeddings (modeling individual galaxies, whose number can vary) and a global embedding (modeling the whole line of sight) in a series of residual blocks.
1
0
3
Second, before samples from the posterior distribution inferred by the surrogate model enter the multi-objective acquisition function (like EHVI), we transform the samples such that they conform to the specified partial ordering of properties.
1
1
2
If you have a strong background in modeling time series with ML, we'd love to chat! Experience with neural differential equations is a plus. 3/5
1
0
2
In the mood for a quick weekend read that packs a punch?📚👊 Check out my translation of a very socially relevant Korean short story, "Blind Spot" by Im Sol-A, published online in the latest issue of Nabillera.
0
0
2
Time series classification with Tensorflow – burakhimmetoglu
0
2
2
AGN variability is believed to be stochastic and commonly modeled as a Gaussian process with a predefined kernel. Latent SDEs (Li et al 2020) offer a more flexible model of the stochastic dynamics.
1
0
2
We then combine the convergence constraints on individual sightlines from the BGNN in a hierarchical Bayesian model via importance weighting to obtain constraints on population-level hyperparameters of the test set.
1
0
2
Monte Carlo Tree Search - beginners guide - Machine learning blog Machine learning blog
0
0
2
😍
New book on graph representation learning by William Hamilton.
PDF --
0
1
10
0
0
2
What is convergence? Say we observe a galaxy in the sky. It may appear to us distorted in shape or magnified/de-magnified in brightness due to gravitational lensing by foreground masses, both luminous and dark. Convergence captures the degree of magnification.
1
0
2
We seek to robustly infer the weak lensing convergence from photometric catalog information (e.g., measured positions, colors) of galaxies along a line of sight.
1
0
2
It demonstrates that lens modeling with Bayesian neural networks (BNN) is accurate and efficient enough so as to enable unbiased recovery of H0 over hundreds (!) of lenses.
1
0
2
Supermassive black holes reside at the centers of most galaxies, feeding on diffuse matter around them. Radiation from matter falling into their gravitational pull makes these galactic centers, or AGNs, some of the most luminous in the Universe.
1
0
2
In all tasks, PropertyDAG-BO identifies significantly more designs that are jointly positive (exceeding a chosen threshold in a given property and all its ancestral properties, like "expressing binders") than does standard BO.
1
1
2
Added to my Thanksgiving reading list...
The book "Foundations of Probabilistic Programming" has now been published, edited by
@alexandra8silva
, Barthe and Katoen, and including the chapter I co-authored with
@schrijvers_tom
, Carroll Morgan and Annabelle McIver. Best of all, it's Open Access!
0
76
266
0
0
2
Upcoming large-sky telescope surveys herald an unprecedented increase in AGN data volume. Our method -- equipped for multiple bands, long gaps of missing data, non-uniform sampling, and random/systematic noise -- promises to accelerate cosmology and AGN physics.
1
0
2
(This modeling choice is well-suited for biological properties, which tend to carry excess zeros and fat tails.)
1
1
2
How to Install Tensorflow GPU with CUDA 9.2 for Python on Ubuntu - 掃文資訊
0
0
2
The BGNN extracts more convergence signal compared to matching number counts, particularly in the high- and low-convergence tails of the training set with sparse samples. Note that number counts matching is limited by sample variance.
1
0
2
The variability patterns of an AGN's light can reveal important information about the physical properties of the underlying black hole, such as its mass and age.
1
0
2
Why do we care about it? Reconstructing a map of convergence across the sky can inform the relationship between luminous matter (which we see in starlight) and dark matter (which we don't see but convergence captures), known as the galaxy-halo connection.
1
0
2
Bonus: We use raytracing in Lenstronomy to enhance the lensing scales of large-scale simulations. Starting with convergence computed on a coarse ~1' grid, we introduce finer fluctuations on galaxy lensing scales of ~1" to generate our training set.
1
1
2
The Real Reason Fashion Companies Like Gap, Zara, Dior, and Chanel Are Now Making Face Masks | Marker
0
0
2
The BNN method of H0 inference only takes ~10 minutes per lens, compared to ~3 hours for the traditional forward modeling method! It promises to become a key tool in exploring ensemble-level systematics in lens modeling for H0 inference ...
1
0
2
I predict a writer will come along and present Seoul to the world
#JamesJoyce
#Dubliners
#Seoulites
#Prophecy
0
1
2
First, we model each objective as zero-inflated, i.e. as a mixture of zeros and a continuous distribution of non-zero values. The surrogate model consists of a binary classifier and a regressor inferring the zeros and non-zeros, respectively.
1
1
2
Opposition to taxation of graduate student tuition waivers and remissions.
0
0
1
How Machine Learning is changing Software Development
0
0
1
LPT: never run `sudo apt-get purge openssl`... but more generally don't get trigger-happy with suggestions on SO (and read the comments!)
0
0
1
Pytorch 코드 작성 팁
0
0
1
WFIRST --> the Nancy Grace Roman Space Telescope. I wouldn't have known about Roman's legacy with Hubble if it weren't for the naming. Loving this trend to honor trailblazing women in astrophysics, following the Vera Rubin Observatory!
0
0
1
@khloek93
my paper's nonexistence saddens me more... I think I'm waiting for it to drop from the sky
0
0
1
If a dark matter halo happens to line up in front of a quasar, we have a *strong* gravitational lens from which we can measure the Hubble constant. The precision of this analysis depends on the total "external" convergence due to objects in the lens environment and line of sight.
1
0
1
The dark side of .io: How the U.K. is making web domain profits from a shady Cold War land deal via
@Gigaom
0
1
1
0
0
1
@SpaceX
What if you added four magnetic helicopters (guardian angels) to help land that rocket? It looks a bit long.
0
0
1
Three mysteries in deep learning: Ensemble, knowledge distillation, and self-distillation - Microsoft Research
0
0
1
The Gentlewoman – Margaret Atwood
0
0
1
... as we look to the Rubin Observatory LSST era when we will discover thousands of lenses!
1
0
1
On a particularly challenging test set with high convergence relative to the training set, BGNN recovers the population mean precisely and without bias, resulting in a sub-percent contribution to the hypothetical error budget on the Hubble constant.
1
0
1
Why Gridcoin Beats Golem Hands Down – Nate On The Net
0
0
0
Seeking the True Story of the Comfort Women | The New Yorker
2
0
1
Editing hosts files if can't access specific websites
0
0
1
Recommended for GNN beginners, with helpful visualizations!
0
0
1
I just bought: The Razer Blade GeForce GTX 1060 VR Ready, 14" HD Gaming Laptop, Intel Core i7 7th Ge... via
@newegg
0
0
1
How artificial intelligence is changing science | Stanford News
0
0
1
Not too far from where I am... surely it's a sign?
The house where Fitzgerald wrote “The Great Gatsby” is for sale and guess what, it’s really nice
http://t.co/AlmYHnvQiH
0
3
5
0
0
1
The Cryptocurrency Funds Have Arrived, And They're Bringing Wall Street Money $COIN $GBTC $ARKK $ARKW
0
0
1
👇
Large-Scale Gravitational Lens Modeling with Bayesian Neural Networks for Accurate and Precise Inference of the Hubble Constant
by
@jiwoncpark
et al.
#NeuralNetwork
#ComputerScience
0
3
8
0
0
1