Ji Won Park

Ji Won Park, Simon Birrer, Madison Ueland et al. · cambridge

We present a Bayesian graph neural network (BGNN) that can estimate the weak lensing convergence ($κ$) from photometric measurements of galaxies along a given line of sight. The method is of particular interest in strong gravitational time delay cosmography (TDC), where characterizing the "external convergence" ($κ_{\rm ext}$) from the lens environment and line of sight is necessary for precise inference of the Hubble constant ($H_0$). Starting from a large-scale simulation with a $κ$ resolution of $\sim$1$'$, we introduce fluctuations on galaxy-galaxy lensing scales of $\sim$1$''$ and extract random sightlines to train our BGNN. We then evaluate the model on test sets with varying degrees of overlap with the training distribution. For each test set of 1,000 sightlines, the BGNN infers the individual $κ$ posteriors, which we combine in a hierarchical Bayesian model to yield constraints on the hyperparameters governing the population. For a test field well sampled by the training set, the BGNN recovers the population mean of $κ$ precisely and without bias, resulting in a contribution to the $H_0$ error budget well under 1\%. In the tails of the training set with sparse samples, the BGNN, which can ingest all available information about each sightline, extracts more $κ$ signal compared to a simplified version of the traditional method based on matching galaxy number counts, which is limited by sample variance. Our hierarchical inference pipeline using BGNNs promises to improve the $κ_{\rm ext}$ characterization for precision TDC. The implementation of our pipeline is available as a public Python package, Node to Joy.

Michael Maser, Ji Won Park, Joshua Yao-Yu Lin et al. · berkeley

We investigate Siamese networks for learning related embeddings for augmented samples of molecular conformers. We find that a non-contrastive (positive-pair only) auxiliary task aids in supervised training of Euclidean neural networks (E3NNs) and increases manifold smoothness (MS) around point-cloud geometries. We demonstrate this property for multiple drug-activity prediction tasks while maintaining relevant performance metrics, and propose an extension of MS to probabilistic and regression settings. We provide an analysis of representation collapse, finding substantial effects of task-weighting, latent dimension, and regularization. We expect the presented protocol to aid in the development of reliable E3NNs from molecular conformers, even for small-data drug discovery programs.

Nathan Ng, Ji Won Park, Jae Hyeon Lee et al.

Biological sequence analysis relies on the ability to denoise the imprecise output of sequencing platforms. We consider a common setting where a short sequence is read out repeatedly using a high-throughput long-read platform to generate multiple subreads, or noisy observations of the same sequence. Denoising these subreads with alignment-based approaches often fails when too few subreads are available or error rates are too high. In this paper, we propose a novel method for blindly denoising sets of sequences without directly observing clean source sequence labels. Our method, Self-Supervised Set Learning (SSSL), gathers subreads together in an embedding space and estimates a single set embedding as the midpoint of the subreads in both the latent and sequence spaces. This set embedding represents the "average" of the subreads and can be decoded into a prediction of the clean sequence. In experiments on simulated long-read DNA data, SSSL methods denoise small reads of $\leq 6$ subreads with 17% fewer errors and large reads of $>6$ subreads with 8% fewer errors compared to the best baseline. On a real dataset of antibody sequences, SSSL improves over baselines on two self-supervised metrics, with a significant improvement on difficult small reads that comprise over 60% of the test set. By accurately denoising these reads, SSSL promises to better realize the potential of high-throughput DNA sequencing data for downstream scientific applications.

Ji Won Park, Samuel Stanton, Saeed Saremi et al.

Bayesian optimization offers a sample-efficient framework for navigating the exploration-exploitation trade-off in the vast design space of biological sequences. Whereas it is possible to optimize the various properties of interest jointly using a multi-objective acquisition function, such as the expected hypervolume improvement (EHVI), this approach does not account for objectives with a hierarchical dependency structure. We consider a common use case where some regions of the Pareto frontier are prioritized over others according to a specified $\textit{partial ordering}$ in the objectives. For instance, when designing antibodies, we would like to maximize the binding affinity to a target antigen only if it can be expressed in live cell culture -- modeling the experimental dependency in which affinity can only be measured for antibodies that can be expressed and thus produced in viable quantities. In general, we may want to confer a partial ordering to the properties such that each property is optimized conditioned on its parent properties satisfying some feasibility condition. To this end, we present PropertyDAG, a framework that operates on top of the traditional multi-objective BO to impose this desired ordering on the objectives, e.g. expression $\rightarrow$ affinity. We demonstrate its performance over multiple simulated active learning iterations on a penicillin production task, toy numerical problem, and a real-world antibody design task.

Drew Prinster, Clara Fannjiang, Ji Won Park et al.

An agent must try new behaviors to explore and improve. In high-stakes environments, an agent that violates safety constraints may cause harm and must be taken offline, curtailing any future interaction. Imitating old behavior is safe, but excessive conservatism discourages exploration. How much behavior change is too much? We show how to use any safe reference policy as a probabilistic regulator for any optimized but untested policy. Conformal calibration on data from the safe policy determines how aggressively the new policy can act, while provably enforcing the user's declared risk tolerance. Unlike conservative optimization methods, we do not assume the user has identified the correct model class nor tuned any hyperparameters. Unlike previous conformal methods, our theory provides finite-sample guarantees even for non-monotonic bounded constraint functions. Our experiments on applications ranging from natural language question answering to biomolecular engineering show that safe exploration is not only possible from the first moment of deployment, but can also improve performance.

Daniel Berenberg, Jae Hyeon Lee, Simon Kelow et al.

Deep generative modeling for biological sequences presents a unique challenge in reconciling the bias-variance trade-off between explicit biological insight and model flexibility. The deep manifold sampler was recently proposed as a means to iteratively sample variable-length protein sequences by exploiting the gradients from a function predictor. We introduce an alternative approach to this guided sampling procedure, multi-segment preserving sampling, that enables the direct inclusion of domain-specific knowledge by designating preserved and non-preserved segments along the input sequence, thereby restricting variation to only select regions. We present its effectiveness in the context of antibody design by training two models: a deep manifold sampler and a GPT-2 language model on nearly six million heavy chain sequences annotated with the IGHV1-18 gene. During sampling, we restrict variation to only the complementarity-determining region 3 (CDR3) of the input. We obtain log probability scores from a GPT-2 model for each sampled CDR3 and demonstrate that multi-segment preserving sampling generates reasonable designs while maintaining the desired, preserved regions.

Ji Won Park, Nataša Tagasovska, Michael Maser et al.

Many scientific and industrial applications require the joint optimization of multiple, potentially competing objectives. Multi-objective Bayesian optimization (MOBO) is a sample-efficient framework for identifying Pareto-optimal solutions. At the heart of MOBO is the acquisition function, which determines the next candidate to evaluate by navigating the best compromises among the objectives. In this paper, we show a natural connection between non-dominated solutions and the extreme quantile of the joint cumulative distribution function (CDF). Motivated by this link, we propose the Pareto-compliant CDF indicator and the associated acquisition function, BOtied. BOtied inherits desirable invariance properties of the CDF, and an efficient implementation with copulas allows it to scale to many objectives. Our experiments on a variety of synthetic and real-world problems demonstrate that BOtied outperforms state-of-the-art MOBO acquisition functions while being computationally efficient for many objectives.

Kyurae Kim, Samuel Gruffaz, Ji Won Park et al.

Simulating the kinetic Langevin dynamics is a popular approach for sampling from distributions, where only their unnormalized densities are available. Various discretizations of the kinetic Langevin dynamics have been considered, where the resulting algorithm is collectively referred to as the kinetic Langevin Monte Carlo (KLMC) or underdamped Langevin Monte Carlo. Specifically, the stochastic exponential Euler discretization, or exponential integrator for short, has previously been studied under strongly log-concave and log-Lipschitz smooth potentials via the synchronous Wasserstein coupling strategy. Existing analyses, however, impose restrictions on the parameters that do not explain the behavior of KLMC under various choices of parameters. In particular, all known results fail to hold in the overdamped regime, suggesting that the exponential integrator degenerates in the overdamped limit. In this work, we revisit the synchronous Wasserstein coupling analysis of KLMC with the exponential integrator. Our refined analysis results in Wasserstein contractions and bounds on the asymptotic bias that hold under weaker restrictions on the parameters, which assert that the exponential integrator is capable of stably simulating the kinetic Langevin dynamics in the overdamped regime, as long as proper time acceleration is applied.

Nataša Tagasovska, Ji Won Park, Matthieu Kirchmeyer et al.

Machine learning (ML) has demonstrated significant promise in accelerating drug design. Active ML-guided optimization of therapeutic molecules typically relies on a surrogate model predicting the target property of interest. The model predictions are used to determine which designs to evaluate in the lab, and the model is updated on the new measurements to inform the next cycle of decisions. A key challenge is that the experimental feedback from each cycle inspires changes in the candidate proposal or experimental protocol for the next cycle, which lead to distribution shifts. To promote robustness to these shifts, we must account for them explicitly in the model training. We apply domain generalization (DG) methods to classify the stability of interactions between an antibody and antigen across five domains defined by design cycles. Our results suggest that foundational models and ensembling improve predictive performance on out-of-distribution domains. We publicly release our codebase extending the DG benchmark ``DomainBed,'' and the associated dataset of antibody sequences and structures emulating distribution shifts across design cycles.

Omid Bazgir, Zichen Wang, Ji Won Park et al.

In anti-cancer drug development, a major scientific challenge is disentangling the complex relationships between high-dimensional genomics data from patient tumor samples, the corresponding tumor's organ of origin, the drug targets associated with given treatments and the resulting treatment response. Furthermore, to realize the aspirations of precision medicine in identifying and adjusting treatments for patients depending on the therapeutic response, there is a need for building tumor dynamic models that can integrate both longitudinal tumor size as well as multimodal, high-content data. In this work, we take a step towards enhancing personalized tumor dynamic predictions by proposing a heterogeneous graph encoder that utilizes a bipartite Graph Convolutional Neural network (GCN) combined with Neural Ordinary Differential Equations (Neural-ODEs). We applied the methodology to a large collection of patient-derived xenograft (PDX) data, spanning a wide variety of treatments (as well as their combinations) on tumors that originated from a number of different organs. We first show that the methodology is able to discover a tumor dynamic model that significantly improves upon an empirical model which is in current use. Additionally, we show that the graph encoder is able to effectively utilize multimodal data to enhance tumor predictions. Our findings indicate that the methodology holds significant promise and offers potential applications in pre-clinical settings.

Ji Won Park, Robert Tibshirani, Kyunghyun Cho

Many risk-sensitive applications require well-calibrated prediction sets over multiple, potentially correlated target variables, for which the prediction algorithm may report correlated errors. In this work, we aim to construct the conformal prediction set accounting for the joint correlation structure of the vector-valued non-conformity scores. Drawing from the rich literature on multivariate quantiles and semiparametric statistics, we propose an algorithm to estimate the $1-α$ quantile of the scores, where $α$ is the user-specified miscoverage rate. In particular, we flexibly estimate the joint cumulative distribution function (CDF) of the scores using nonparametric vine copulas and improve the asymptotic efficiency of the quantile estimate using its influence function. The vine decomposition allows our method to scale well to a large number of targets. As well as guaranteeing asymptotically exact coverage, our method yields desired coverage and competitive efficiency on a range of real-world regression problems, including those with missing-at-random labels in the calibration set.

Andor Vári-Kakas, Ji Won Park, Natasa Tagasovska

Aligning large language models (LLMs) to desirable human values requires balancing multiple, potentially conflicting objectives such as helpfulness, truthfulness, and harmlessness, which presents a multi-objective optimisation challenge. Most alignment pipelines rely on a fixed scalarisation of these objectives, which can introduce procedural unfairness by systematically under-weighting harder-to-optimise or minority objectives. To promote more equitable trade-offs, we introduce MGDA-Decoupled, a geometry-based multi-objective optimisation algorithm that finds a shared descent direction while explicitly accounting for each objective's convergence dynamics. In contrast to prior methods that depend on reinforcement learning (e.g., GAPO) or explicit reward models (e.g., MODPO), our approach operates entirely within the lightweight Direct Preference Optimisation (DPO) paradigm. Experiments on the UltraFeedback dataset show that geometry-aware methods -- and MGDA-Decoupled in particular -- achieve the highest win rates against golden responses, both overall and per objective.

Taro Makino, Ji Won Park, Natasa Tagasovska et al.

Real-world datasets often combine data collected under different experimental conditions. This yields larger datasets, but also introduces spurious correlations that make it difficult to model the phenomena of interest. We address this by learning two embeddings to independently represent the phenomena of interest and the spurious correlations. The embedding representing the phenomena of interest is correlated with the target variable $y$, and is invariant to the environment variable $e$. In contrast, the embedding representing the spurious correlations is correlated with $e$. The invariance to $e$ is difficult to achieve on real-world datasets. Our primary contribution is an algorithm called Supervised Contrastive Block Disentanglement (SCBD) that effectively enforces this invariance. It is based purely on Supervised Contrastive Learning, and applies to real-world data better than existing approaches. We empirically validate SCBD on two challenging problems. The first problem is domain generalization, where we achieve strong performance on a synthetic dataset, as well as on Camelyon17-WILDS. We introduce a single hyperparameter $α$ to control the degree of invariance to $e$. When we increase $α$ to strengthen the degree of invariance, out-of-distribution performance improves at the expense of in-distribution performance. The second problem is batch correction, in which we apply SCBD to preserve biological signal and remove inter-well batch effects when modeling single-cell perturbations from 26 million Optical Pooled Screening images.

Ji Won Park, Kyunghyun Cho

Accurately estimating semantic aleatoric and epistemic uncertainties in large language models (LLMs) is particularly challenging in free-form question answering (QA), where obtaining stable estimates often requires many expensive generations. We introduce a diversity-steered sampler that discourages semantically redundant outputs during decoding, covers both autoregressive and masked diffusion paradigms, and yields substantial sample-efficiency gains. The key idea is to inject a continuous semantic-similarity penalty into the model's proposal distribution using a natural language inference (NLI) model lightly finetuned on partial prefixes or intermediate diffusion states. We debias downstream uncertainty estimates with importance reweighting and shrink their variance with control variates. Across four QA benchmarks, our method matches or surpasses baselines while covering more semantic clusters with the same number of samples. Being modular and requiring no gradient access to the base LLM, the framework promises to serve as a drop-in enhancement for uncertainty estimation in risk-sensitive model deployments.

Clara Fannjiang, Ji Won Park

Algorithms for machine learning-guided design, or design algorithms, use machine learning-based predictions to propose novel objects with desired property values. Given a new design task -- for example, to design novel proteins with high binding affinity to a therapeutic target -- one must choose a design algorithm and specify any hyperparameters and predictive and/or generative models involved. How can these decisions be made such that the resulting designs are successful? This paper proposes a method for design algorithm selection, which aims to select design algorithms that will produce a distribution of design labels satisfying a user-specified success criterion -- for example, that at least ten percent of designs' labels exceed a threshold. It does so by combining designs' predicted property values with held-out labeled data to reliably forecast characteristics of the label distributions produced by different design algorithms, building upon techniques from prediction-powered inference. The method is guaranteed with high probability to return design algorithms that yield successful label distributions (or the null set if none exist), if the density ratios between the design and labeled data distributions are known. We demonstrate the method's effectiveness in simulated protein and RNA design tasks, in settings with either known or estimated density ratios.

Saeed Saremi, Ji Won Park, Francis Bach

We introduce a theoretical framework for sampling from unnormalized densities based on a smoothing scheme that uses an isotropic Gaussian kernel with a single fixed noise scale. We prove one can decompose sampling from a density (minimal assumptions made on the density) into a sequence of sampling from log-concave conditional densities via accumulation of noisy measurements with equal noise levels. Our construction is unique in that it keeps track of a history of samples, making it non-Markovian as a whole, but it is lightweight algorithmically as the history only shows up in the form of a running empirical mean of samples. Our sampling algorithm generalizes walk-jump sampling (Saremi & Hyvärinen, 2019). The "walk" phase becomes a (non-Markovian) chain of (log-concave) Markov chains. The "jump" from the accumulated measurements is obtained by empirical Bayes. We study our sampling algorithm quantitatively using the 2-Wasserstein metric and compare it with various Langevin MCMC algorithms. We also report a remarkable capacity of our algorithm to "tunnel" between modes of a distribution.

Ji Won Park, Ashley Villar, Yin Li et al.

Among the most extreme objects in the Universe, active galactic nuclei (AGN) are luminous centers of galaxies where a black hole feeds on surrounding matter. The variability patterns of the light emitted by an AGN contain information about the physical properties of the underlying black hole. Upcoming telescopes will observe over 100 million AGN in multiple broadband wavelengths, yielding a large sample of multivariate time series with long gaps and irregular sampling. We present a method that reconstructs the AGN time series and simultaneously infers the posterior probability density distribution (PDF) over the physical quantities of the black hole, including its mass and luminosity. We apply this method to a simulated dataset of 11,000 AGN and report precision and accuracy of 0.4 dex and 0.3 dex in the inferred black hole mass. This work is the first to address probabilistic time series reconstruction and parameter inference for AGN in an end-to-end fashion.

Ji Won Park, Sebastian Wagner-Carena, Simon Birrer et al.

We investigate the use of approximate Bayesian neural networks (BNNs) in modeling hundreds of time-delay gravitational lenses for Hubble constant ($H_0$) determination. Our BNN was trained on synthetic HST-quality images of strongly lensed active galactic nuclei (AGN) with lens galaxy light included. The BNN can accurately characterize the posterior PDFs of model parameters governing the elliptical power-law mass profile in an external shear field. We then propagate the BNN-inferred posterior PDFs into ensemble $H_0$ inference, using simulated time delay measurements from a plausible dedicated monitoring campaign. Assuming well-measured time delays and a reasonable set of priors on the environment of the lens, we achieve a median precision of $9.3$\% per lens in the inferred $H_0$. A simple combination of 200 test-set lenses results in a precision of 0.5 $\textrm{km s}^{-1} \textrm{ Mpc}^{-1}$ ($0.7\%$), with no detectable bias in this $H_0$ recovery test. The computation time for the entire pipeline -- including the training set generation, BNN training, and $H_0$ inference -- translates to 9 minutes per lens on average for 200 lenses and converges to 6 minutes per lens as the sample size is increased. Being fully automated and efficient, our pipeline is a promising tool for exploring ensemble-level systematics in lens modeling for $H_0$ inference.

Sebastian Wagner-Carena, Ji Won Park, Simon Birrer et al.

In the past few years, approximate Bayesian Neural Networks (BNNs) have demonstrated the ability to produce statistically consistent posteriors on a wide range of inference problems at unprecedented speed and scale. However, any disconnect between training sets and the distribution of real-world objects can introduce bias when BNNs are applied to data. This is a common challenge in astrophysics and cosmology, where the unknown distribution of objects in our Universe is often the science goal. In this work, we incorporate BNNs with flexible posterior parameterizations into a hierarchical inference framework that allows for the reconstruction of population hyperparameters and removes the bias introduced by the training distribution. We focus on the challenge of producing posterior PDFs for strong gravitational lens mass model parameters given Hubble Space Telescope (HST) quality single-filter, lens-subtracted, synthetic imaging data. We show that the posterior PDFs are sufficiently accurate (i.e., statistically consistent with the truth) across a wide variety of power-law elliptical lens mass distributions. We then apply our approach to test data sets whose lens parameters are drawn from distributions that are drastically different from the training set. We show that our hierarchical inference framework mitigates the bias introduced by an unrepresentative training set's interim prior. Simultaneously, given a sufficiently broad training set, we can precisely reconstruct the population hyperparameters governing our test distributions. Our full pipeline, from training to hierarchical inference on thousands of lenses, can be run in a day. The framework presented here will allow us to efficiently exploit the full constraining power of future ground- and space-based surveys.