Search

Despite significant advancements in automatic brain tumor segmentation methods, their performance is not guaranteed when certain MR sequences are missing. Addressing this issue, it is crucial to synthesize the missing MR images that reflect the unique characteristics of the absent modality with precise tumor representation. Typically, MRI synthesis methods generate partial images rather than full-sized volumes due to computational constraints. This limitation can lead to a lack of comprehensive 3D volumetric information and result in image artifacts during the merging process. In this paper, we propose a two-stage approach that first synthesizes MR images from 2D slices using a novel intensity encoding method and then refines the synthesized MRI. The proposed intensity encoding reduces artifacts when synthesizing MRI on a 2D slice basis. Then, the \textit{Refiner}, which leverages complete 3D volume information, further improves the quality of the synthesized images and enhances their applicability to segmentation methods. Experimental results demonstrate that the intensity encoding effectively minimizes artifacts in the synthesized MRI and improves perceptual quality. Furthermore, using the \textit{Refiner} on synthesized MRI significantly improves brain tumor segmentation results, highlighting the potential of our approach in practical applications., Comment: MICCAI 2024 BraSyn Challenge 1st place

We introduce GENomic Encoding REpresentation with Language Model (GENEREL), a framework designed to bridge genetic and biomedical knowledge bases. What sets GENEREL apart is its ability to fine-tune language models to infuse biological knowledge behind clinical concepts such as diseases and medications. This fine-tuning enables the model to capture complex biomedical relationships more effectively, enriching the understanding of how genomic data connects to clinical outcomes. By constructing a unified embedding space for biomedical concepts and a wide range of common SNPs from sources such as patient-level data, biomedical knowledge graphs, and GWAS summaries, GENEREL aligns the embeddings of SNPs and clinical concepts through multi-task contrastive learning. This allows the model to adapt to diverse natural language representations of biomedical concepts while bypassing the limitations of traditional code mapping systems across different data sources. Our experiments demonstrate GENEREL's ability to effectively capture the nuanced relationships between SNPs and clinical concepts. GENEREL also emerges to discern the degree of relatedness, potentially allowing for a more refined identification of concepts. This pioneering approach in constructing a unified embedding system for both SNPs and biomedical concepts enhances the potential for data integration and discovery in biomedical research., Comment: 15 pages, 2 figures, 5 tables

Contour maps are an essential tool for exploring spatial features of the terrain, such as distance, directions, and surface gradient among the contour areas. User interactions in contour-based visualizations create approaches to visual analysis that are noticeably different from the perspective of human cognition. As such, various interactive approaches have been introduced to improve system usability and enhance human cognition for complex and large-scale spatial data exploration. However, what user interaction means for contour maps, its purpose, when to leverage, and design primitives have yet to be investigated in the context of analysis tasks. Therefore, further research is needed to better understand and quantify the potentials and benefits offered by user interactions in contour-based geospatial visualizations designed to support analytical tasks. In this paper, we present a contour-based interactive geospatial visualization designed for analytical tasks. We conducted a crowd-sourced user study (N=62) to examine the impact of interactive features on analysis using contour-based geospatial visualizations. Our results show that the interactive features aid in their data analysis and understanding in terms of spatial data extent, map layout, task complexity, and user expertise. Finally, we discuss our findings in-depth, which will serve as guidelines for future design and implementation of interactive features in support of case-specific analytical tasks on contour-based geospatial views.

Dynamical systems form the foundation of scientific discovery, traditionally modeled with predefined state variables such as the angle and angular velocity, and differential equations such as the equation of motion for a single pendulum. We propose an approach to discover a set of state variables that preserve the smoothness of the system dynamics and to construct a vector field representing the system's dynamics equation, automatically from video streams without prior physical knowledge. The prominence and effectiveness of the proposed approach are demonstrated through both quantitative and qualitative analyses of various dynamical systems, including the prediction of characteristic frequencies and the identification of chaotic and limit cycle behaviors. This shows the potential of our approach to assist human scientists in scientific discovery.

Large Language Models (LLMs) have exhibited remarkable capabilities in many complex tasks including mathematical reasoning. However, traditional approaches heavily rely on ensuring self-consistency within single prompting method, which limits the exploration of diverse problem-solving strategies. This study addresses these limitations by performing an experimental analysis of distinct prompting methods within the domain of mathematical reasoning. Our findings demonstrate that each method explores a distinct search space, and this differentiation becomes more evident with increasing problem complexity. To leverage this phenomenon, we applied efficient sampling process that uniformly combines samples from these diverse methods, which not only expands the maximum search space but achieves higher performance with fewer runs compared to single methods. Especially, within the subset of difficult questions of MATH dataset named MATH-hard, The maximum search space was achieved while utilizing approximately 43% fewer runs than single methods on average. These findings highlight the importance of integrating diverse problem-solving strategies to enhance the reasoning abilities of LLMs., Comment: 6 pages, 4 figures

In this paper, we study the significance of ecological interactions and separation of birth and death dynamics in heterogeneous stochastic populations. Interactions can manifest through the birth dynamics, the death dynamics, or some combination of the two. The underlying mechanisms are important but often implicit in data. We propose an inference method for disambiguating the types of interaction and the birth and death processes from time series data of a stochastic n-type heterogeneous population. The interspecies interactions considered can be competitive, antagonistic, or mutualistic. The inference method is then validated in the example of two-type Lotka-Volterra dynamics. Utilizing stochastic fluctuations enables us to estimate additional parameters in a stochastic Lotka-Volterra model, which are not identifiable in a deterministic model. We also show that different pairs of birth and death rates with the same net growth rate manifest different time series statistics and survival probabilities., Comment: 31 pages, 7 figures

With the rapid growth in the use of fine-tuning for large language models (LLMs), optimizing fine-tuning while keeping inference efficient has become highly important. However, this is a challenging task as it requires improvements in all aspects, including inference speed, fine-tuning speed, memory consumption, and, most importantly, model quality. Previous studies have attempted to achieve this by combining quantization with fine-tuning, but they have failed to enhance all four aspects simultaneously. In this study, we propose a new lightweight technique called Quantization for Efficient Fine-Tuning (QEFT). QEFT accelerates both inference and fine-tuning, is supported by robust theoretical foundations, offers high flexibility, and maintains good hardware compatibility. Our extensive experiments demonstrate that QEFT matches the quality and versatility of full-precision parameter-efficient fine-tuning, while using fewer resources. Our code is available at https://github.com/xvyaward/qeft., Comment: Accepted at Findings of EMNLP 2024

This paper proposes incentive mechanisms that promote eco-driving in transportation networks with the over-arching objective of minimizing emissions. The transportation system operator provides the drivers with energy-efficient driving guidance throughout their trips, and their eco-driving levels are measured by how closely they follow this guidance via vehicle telematics. Drivers choose their eco-driving levels to optimize a combination of their travel times and their emissions. To obtain optimal budget allocation and recommendations for the incentive mechanism, the system operator gathers drivers' preferences, or types, to assess each driver's trip urgency and natural willingness to eco-drive. In a setting where drivers truthfully report their types, we introduce the first-best incentive mechanism and show that the obedience condition holds (i.e., drivers find it optimal to comply with the system operator's recommendations) when the recommended eco-driving profile constitutes a Nash equilibrium. Moreover, in a setting where drivers can strategically report their types, we introduce the second-best incentive mechanism and show that the proposed mechanism is incentive-compatible (i.e., drivers find it optimal to be truthful). Under this mechanism, we also show that all equilibrium outcomes are at least as good as the recommended eco-driving profile in terms of the system operator's objective. Overall, this work offers a framework for designing eco-driving incentive mechanisms while considering both the strategic behavior of individual drivers and the network effects of collective decision-making.

With the recent growth of video-based Social Network Service (SNS) platforms, the demand for video editing among common users has increased. However, video editing can be challenging due to the temporally-varying factors such as camera movement and moving objects. While modern atlas-based video editing methods have addressed these issues, they often fail to edit videos including complex motion or multiple moving objects, and demand excessive computational cost, even for very simple edits. In this paper, we propose a novel region-of-interest (ROI)-based video editing framework: ROI-based Neural Atlas (RNA). Unlike prior work, RNA allows users to specify editing regions, simplifying the editing process by removing the need for foreground separation and atlas modeling for foreground objects. However, this simplification presents a unique challenge: acquiring a mask that effectively handles occlusions in the edited area caused by moving objects, without relying on an additional segmentation model. To tackle this, we propose a novel mask refinement approach designed for this specific challenge. Moreover, we introduce a soft neural atlas model for video reconstruction to ensure high-quality editing results. Extensive experiments show that RNA offers a more practical and efficient editing solution, applicable to a wider range of videos with superior quality compared to prior methods., Comment: ACCV2024

The first very long baseline interferometry (VLBI) detections at 870$\mu$m wavelength (345$\,$GHz frequency) are reported, achieving the highest diffraction-limited angular resolution yet obtained from the surface of the Earth, and the highest-frequency example of the VLBI technique to date. These include strong detections for multiple sources observed on inter-continental baselines between telescopes in Chile, Hawaii, and Spain, obtained during observations in October 2018. The longest-baseline detections approach 11$\,$G$\lambda$ corresponding to an angular resolution, or fringe spacing, of 19$\mu$as. The Allan deviation of the visibility phase at 870$\mu$m is comparable to that at 1.3$\,$mm on the relevant integration time scales between 2 and 100$\,$s. The detections confirm that the sensitivity and signal chain stability of stations in the Event Horizon Telescope (EHT) array are suitable for VLBI observations at 870$\mu$m. Operation at this short wavelength, combined with anticipated enhancements of the EHT, will lead to a unique high angular resolution instrument for black hole studies, capable of resolving the event horizons of supermassive black holes in both space and time., Comment: Corresponding author: S. Doeleman

Syllables are compositional units of spoken language that play a crucial role in human speech perception and production. However, current neural speech representations lack structure, resulting in dense token sequences that are costly to process. To bridge this gap, we propose a new model, Sylber, that produces speech representations with clean and robust syllabic structure. Specifically, we propose a self-supervised model that regresses features on syllabic segments distilled from a teacher model which is an exponential moving average of the model in training. This results in a highly structured representation of speech features, offering three key benefits: 1) a fast, linear-time syllable segmentation algorithm, 2) efficient syllabic tokenization with an average of 4.27 tokens per second, and 3) syllabic units better suited for lexical and syntactic understanding. We also train token-to-speech generative models with our syllabic units and show that fully intelligible speech can be reconstructed from these tokens. Lastly, we observe that categorical perception, a linguistic phenomenon of speech perception, emerges naturally in our model, making the embedding space more categorical and sparse than previous self-supervised learning approaches. Together, we present a novel self-supervised approach for representing speech as syllables, with significant potential for efficient speech tokenization and spoken language modeling.

Large language models (LLMs), like ChatGPT, have shown that even trained with noisy prior data, they can generalize effectively to new tasks through in-context learning (ICL) and pre-training techniques. Motivated by this, we explore whether a similar approach can be applied to scientific foundation models (SFMs). Our methodology is structured as follows: (i) we collect low-cost physics-informed neural network (PINN)-based approximated prior data in the form of solutions to partial differential equations (PDEs) constructed through an arbitrary linear combination of mathematical dictionaries; (ii) we utilize Transformer architectures with self and cross-attention mechanisms to predict PDE solutions without knowledge of the governing equations in a zero-shot setting; (iii) we provide experimental evidence on the one-dimensional convection-diffusion-reaction equation, which demonstrate that pre-training remains robust even with approximated prior data, with only marginal impacts on test accuracy. Notably, this finding opens the path to pre-training SFMs with realistic, low-cost data instead of (or in conjunction with) numerical high-cost data. These results support the conjecture that SFMs can improve in a manner similar to LLMs, where fully cleaning the vast set of sentences crawled from the Internet is nearly impossible.

The process of creating training data to teach models is currently driven by humans, who manually analyze model weaknesses and plan how to create data that improves a student model. Recent approaches using LLMs as annotators reduce human effort, but still require humans to interpret feedback from evaluations and control the LLM to produce data the student needs. Automating this labor-intensive process by creating autonomous data generation agents - or teachers - is desirable, but requires environments that can simulate the feedback-driven, iterative, closed loop of data creation. To enable rapid and scalable testing for such agents and their modules, we introduce DataEnvGym, a testbed of teacher environments for data generation agents. DataEnvGym frames data generation as a sequential decision-making task, involving an agent consisting of a data generation policy (which generates a plan for creating training data) and a data generation engine (which transforms the plan into data), inside an environment that provides student feedback. The agent's goal is to improve student performance. Students are iteratively trained and evaluated on generated data, with their feedback (in the form of errors or weak skills) being reported to the agent after each iteration. DataEnvGym includes multiple teacher environment instantiations across 3 levels of structure in the state representation and action space. More structured environments are based on inferred skills and offer more interpretability and curriculum control. We support 3 diverse tasks (math, code, and VQA) and test multiple students and teachers. Example agents in our teaching environments can iteratively improve students across tasks and settings. Moreover, we show that environments teach different skill levels and test variants of key modules, pointing to future work in improving data generation agents, engines, and feedback mechanisms., Comment: Project Page: https://DataEnvGym.github.io

As training datasets grow larger, we aspire to develop models that generalize well to any diverse test distribution, even if the latter deviates significantly from the training data. Various approaches like domain adaptation, domain generalization, and robust optimization attempt to address the out-of-distribution challenge by posing assumptions about the relation between training and test distribution. Differently, we adopt a more conservative perspective by accounting for the worst-case error across all sufficiently diverse test distributions within a known domain. Our first finding is that training on a uniform distribution over this domain is optimal. We also interrogate practical remedies when uniform samples are unavailable by considering methods for mitigating non-uniformity through finetuning and rebalancing. Our theory provides a mathematical grounding for previous observations on the role of entropy and rebalancing for o.o.d. generalization and foundation model training. We also provide new empirical evidence across tasks involving o.o.d. shifts which illustrate the broad applicability of our perspective.

In this paper, we address the challenge of speech enhancement in real-world recordings, which often contain various forms of distortion, such as background noise, reverberation, and microphone artifacts. We revisit the use of Generative Adversarial Networks (GANs) for speech enhancement and theoretically show that GANs are naturally inclined to seek the point of maximum density within the conditional clean speech distribution, which, as we argue, is essential for the speech enhancement task. We study various feature extractors for perceptual loss to facilitate the stability of adversarial training, developing a methodology for probing the structure of the feature space. This leads us to integrate WavLM-based perceptual loss into MS-STFT adversarial training pipeline, creating an effective and stable training procedure for the speech enhancement model. The resulting speech enhancement model, which we refer to as FINALLY, builds upon the HiFi++ architecture, augmented with a WavLM encoder and a novel training pipeline. Empirical results on various datasets confirm our model's ability to produce clear, high-quality speech at 48 kHz, achieving state-of-the-art performance in the field of speech enhancement. Demo page: https://samsunglabs.github.io/FINALLY-page, Comment: Accepted to NeurIPS 2024

In recent years, there has been considerable interest in designing stochastic first-order algorithms to tackle finite-sum smooth minimax problems. To obtain the gradient estimates, one typically relies on the uniform sampling-with-replacement scheme or various sampling-without-replacement (also known as shuffling) schemes. While the former is easier to analyze, the latter often have better empirical performance. In this paper, we propose a novel single-loop stochastic gradient descent-ascent (GDA) algorithm that employs both shuffling schemes and variance reduction to solve nonconvex-strongly concave smooth minimax problems. We show that the proposed algorithm achieves $\epsilon$-stationarity in expectation in $\mathcal{O}(\kappa^2 \epsilon^{-2})$ iterations, where $\kappa$ is the condition number of the problem. This outperforms existing shuffling schemes and matches the complexity of the best-known sampling-with-replacement algorithms. Our proposed algorithm also achieves the same complexity as that of its deterministic counterpart, the two-timescale GDA algorithm. Our numerical experiments demonstrate the superior performance of the proposed algorithm.

In-context learning (ICL) describes a language model's ability to generate outputs based on a set of input demonstrations and a subsequent query. To understand this remarkable capability, researchers have studied simplified, stylized models. These studies have consistently observed long loss plateaus, during which models exhibit minimal improvement, followed by a sudden, rapid surge of learning. In this work, we reveal that training on multiple diverse ICL tasks simultaneously shortens the loss plateaus, making each task easier to learn. This finding is surprising as it contradicts the natural intuition that the combined complexity of multiple ICL tasks would lengthen the learning process, not shorten it. Our result suggests that the recent success in large-scale training of language models may be attributed not only to the richness of the data at scale but also to the easier optimization (training) induced by the diversity of natural language training data.

In this paper, we propose a new method to enhance compositional understanding in pre-trained vision and language models (VLMs) without sacrificing performance in zero-shot multi-modal tasks. Traditional fine-tuning approaches often improve compositional reasoning at the cost of degrading multi-modal capabilities, primarily due to the use of global hard negative (HN) loss, which contrasts global representations of images and texts. This global HN loss pushes HN texts that are highly similar to the original ones, damaging the model's multi-modal representations. To overcome this limitation, we propose Fine-grained Selective Calibrated CLIP (FSC-CLIP), which integrates local hard negative loss and selective calibrated regularization. These innovations provide fine-grained negative supervision while preserving the model's representational integrity. Our extensive evaluations across diverse benchmarks for both compositionality and multi-modal tasks show that FSC-CLIP not only achieves compositionality on par with state-of-the-art models but also retains strong multi-modal capabilities. Code is available at: https://github.com/ytaek-oh/fsc-clip., Comment: EMNLP 2024 (Long, Main). Project page: https://ytaek-oh.github.io/fsc-clip

This paper studies the nonsmooth optimization landscape of the $\ell_1$-norm rank-one symmetric matrix factorization problem using tools from second-order variational analysis. Specifically, as the main finding of this paper, we show that any second-order stationary point (and thus local minimizer) of the problem is actually globally optimal. Besides, some other results concerning the landscape of the problem, such as a complete characterization of the set of stationary points, are also developed, which should be interesting in their own rights. Furthermore, with the above theories, we revisit existing results on the generic minimizing behavior of simple algorithms for nonsmooth optimization and showcase the potential risk of their applications to our problem through several examples. Our techniques can potentially be applied to analyze the optimization landscapes of a variety of other more sophisticated nonsmooth learning problems, such as robust low-rank matrix recovery.

In this paper, we study equality-type Clarke subdifferential chain rules of matrix factorization and factorization machine. Specifically, we show for these problems that provided the latent dimension is larger than some multiple of the problem size (i.e., slightly overparameterized) and the loss function is locally Lipschitz, the subdifferential chain rules hold everywhere. In addition, we examine the tightness of the analysis through some interesting constructions and make some important observations from the perspective of optimization; e.g., we show that for all this type of problems, computing a stationary point is trivial. Some tensor generalizations and neural extensions are also discussed, albeit they remain mostly open.

In-context Learning (ICL) is an emerging few-shot learning paradigm on Language Models (LMs) with inner mechanisms un-explored. There are already existing works describing the inner processing of ICL, while they struggle to capture all the inference phenomena in large language models. Therefore, this paper proposes a comprehensive circuit to model the inference dynamics and try to explain the observed phenomena of ICL. In detail, we divide ICL inference into 3 major operations: (1) Summarize: LMs encode every input text (demonstrations and queries) into linear representation in the hidden states with sufficient information to solve ICL tasks. (2) Semantics Merge: LMs merge the encoded representations of demonstrations with their corresponding label tokens to produce joint representations of labels and demonstrations. (3) Feature Retrieval and Copy: LMs search the joint representations similar to the query representation on a task subspace, and copy the searched representations into the query. Then, language model heads capture these copied label representations to a certain extent and decode them into predicted labels. The proposed inference circuit successfully captured many phenomena observed during the ICL process, making it a comprehensive and practical explanation of the ICL inference process. Moreover, ablation analysis by disabling the proposed steps seriously damages the ICL performance, suggesting the proposed inference circuit is a dominating mechanism. Additionally, we confirm and list some bypass mechanisms that solve ICL tasks in parallel with the proposed circuit., Comment: 31 pages, 37 figures, 6 tables, ICLR 2025 under review

We implement a bootstrap method that combines Schwinger-Dyson equations, thermal inequalities, and semidefinite relaxations of matrix logarithm in the ungauged one-matrix quantum mechanics, at finite rank N as well as in the large N limit, and determine finite temperature observables that interpolate between available analytic results in the low and high temperature limits respectively. We also obtain bootstrap bounds on thermal phase transition as well as preliminary results in the ungauged two-matrix quantum mechanics., Comment: 31 pages, 8 figures

Trolling in online communities typically involves disruptive behaviors such as provoking anger and manipulating discussions, leading to a polarized atmosphere and emotional distress. Robust moderation is essential for mitigating these negative impacts and maintaining a healthy and constructive community atmosphere. However, effectively addressing trolls is difficult because their behaviors vary widely and require different response strategies (RSs) to counter them. This diversity makes it challenging to choose an appropriate RS for each specific situation. To address this challenge, our research investigates whether humans have preferred strategies tailored to different types of trolling behaviors. Our findings reveal a correlation between the types of trolling encountered and the preferred RS. In this paper, we introduce a methodology for generating counter-responses to trolls by recommending appropriate RSs, supported by a dataset aligning these strategies with human preferences across various troll contexts. The experimental results demonstrate that our proposed approach guides constructive discussion and reduces the negative effects of trolls, thereby enhancing the online community environment., Comment: Findings of EMNLP 2024

Competitive equilibrium (CE) for chores has recently attracted significant attention, with many algorithms proposed to approximately compute it. However, existing algorithms either lack iterate convergence guarantees to an exact CE or require solving high-dimensional linear or quadratic programming subproblems. This paper overcomes these issues by proposing a novel unconstrained difference-of-convex formulation, whose stationary points correspond precisely to the CE for chores. We show that the new formulation possesses the local error bound property and the Kurdyka-{\L}ojasiewicz property with an exponent of $1/2$. Consequently, we present the first algorithm whose iterates provably converge linearly to an exact CE for chores. Furthermore, by exploiting the max structure within our formulation and applying smoothing techniques, we develop a subproblem-free algorithm that finds an approximate CE in polynomial time. Numerical experiments demonstrate that the proposed algorithms outperform the state-of-the-art method., Comment: Accepted by WINE 2024

We introduce Sparse Physics Informed Backpropagation (SPInProp), a new class of methods for accelerating backpropagation for a specialized neural network architecture called Low Rank Neural Representation (LRNR). The approach exploits the low rank structure within LRNR and constructs a reduced neural network approximation that is much smaller in size. We call the smaller network FastLRNR. We show that backpropagation of FastLRNR can be substituted for that of LRNR, enabling a significant reduction in complexity. We apply SPInProp to a physics informed neural networks framework and demonstrate how the solution of parametrized partial differential equations is accelerated., Comment: 10 pages, 3 figures

Micromechanical resonators are essential components in time-keeping and sensing devices due to their high frequency, high quality factor, and sensitivity. However, their extremely low damping can lead to various nonlinear phenomena that can compromise frequency stability. A major limiting factor is the Duffing hardening effect, which causes frequency drift through amplitude variations, known as the amplitude-frequency effect. Recently, internal resonance (InRes) has emerged as an effective approach to mitigate this issue and enhance frequency stabilization. In this study, we investigate the frequency stabilization mechanisms of 1:2 and 1:3 InRes using a generalized two-mode reduced-order model that includes Duffing nonlinearity and nonlinear modal coupling. By analyzing the frequency response curves and pi/2-backbone curves, we demonstrate how different parameters affect the effectiveness of frequency stabilization. Our results identify two distinct regimes depending on the coupling strength relative to the stiffening effect as a key factor in determining the stabilization mechanism. For the regime of weak coupling, both 1:2 and 1:3 InRes achieve frequency stabilization through amplitude and frequency saturation over a range of forcing amplitudes. In contrast, strong coupling reduces the amplitude-frequency effect by forming an asymptote line for 1:2 InRes or a zero-dispersion point for 1:3 InRes. These insights offer valuable guidelines for designing micromechanical resonators with high-frequency stability, highlighting InRes as a robust tool for enhancing performance in practical applications., Comment: 31 pages, 7 pages

The paper proposes a moving sum methodology for detecting multiple change points in high-dimensional time series under a factor model, where changes are attributed to those in loadings as well as emergence or disappearance of factors. We establish the asymptotic null distribution of the proposed test for family-wise error control, and show the consistency of the procedure for multiple change point estimation. Simulation studies and an application to a large dataset of volatilities demonstrate the competitive performance of the proposed method.

In this work, we investigate how a model's tendency to broadly integrate its parametric knowledge evolves throughout pretraining, and how this behavior affects overall performance, particularly in terms of knowledge acquisition and forgetting. We introduce the concept of knowledge entropy, which quantifies the range of memory sources the model engages with; high knowledge entropy indicates that the model utilizes a wide range of memory sources, while low knowledge entropy suggests reliance on specific sources with greater certainty. Our analysis reveals a consistent decline in knowledge entropy as pretraining advances. We also find that the decline is closely associated with a reduction in the model's ability to acquire and retain knowledge, leading us to conclude that diminishing knowledge entropy (smaller number of active memory sources) impairs the model's knowledge acquisition and retention capabilities. We find further support for this by demonstrating that increasing the activity of inactive memory sources enhances the model's capacity for knowledge acquisition and retention.

Place recognition plays an important role in achieving robust long-term autonomy. Real-world robots face a wide range of weather conditions (e.g. overcast, heavy rain, and snowing) and most sensors (i.e. camera, LiDAR) essentially functioning within or near-visible electromagnetic waves are sensitive to adverse weather conditions, making reliable localization difficult. In contrast, radar is gaining traction due to long electromagnetic waves, which are less affected by environmental changes and weather independence. In this work, we propose a radar-based lightweight and robust place recognition. We achieve rotational invariance and lightweight by selecting a one-dimensional ring-shaped description and robustness by mitigating the impact of false detection utilizing opposite noise characteristics between free space and feature. In addition, the initial heading can be estimated, which can assist in building a SLAM pipeline that combines odometry and registration, which takes into account onboard computing. The proposed method was tested for rigorous validation across various scenarios (i.e. single session, multi-session, and different weather conditions). In particular, we validate our descriptor achieving reliable place recognition performance through the results of extreme environments that lacked structural information such as an OORD dataset., Comment: 8 pages, 8 figures, accepted to RA-L

Real-life robot navigation involves more than just reaching a destination; it requires optimizing movements while addressing scenario-specific goals. An intuitive way for humans to express these goals is through abstract cues like verbal commands or rough sketches. Such human guidance may lack details or be noisy. Nonetheless, we expect robots to navigate as intended. For robots to interpret and execute these abstract instructions in line with human expectations, they must share a common understanding of basic navigation concepts with humans. To this end, we introduce CANVAS, a novel framework that combines visual and linguistic instructions for commonsense-aware navigation. Its success is driven by imitation learning, enabling the robot to learn from human navigation behavior. We present COMMAND, a comprehensive dataset with human-annotated navigation results, spanning over 48 hours and 219 km, designed to train commonsense-aware navigation systems in simulated environments. Our experiments show that CANVAS outperforms the strong rule-based system ROS NavStack across all environments, demonstrating superior performance with noisy instructions. Notably, in the orchard environment, where ROS NavStack records a 0% total success rate, CANVAS achieves a total success rate of 67%. CANVAS also closely aligns with human demonstrations and commonsense constraints, even in unseen environments. Furthermore, real-world deployment of CANVAS showcases impressive Sim2Real transfer with a total success rate of 69%, highlighting the potential of learning from human demonstrations in simulated environments for real-world applications., Comment: project page https://worv-ai.github.io/canvas

Multi-Q magnetic structures on triangular lattices, with their two-dimensional topological spin texture, have attracted significant interest. However, unambiguously confirming their formation by excluding the presence of three equally-populated single-Q domains remains challenging. In the metallic triangular lattice antiferromagnet Co1/3TaS2, two magnetic ground states have been suggested at different temperature ranges, with the low-temperature phase being a triple-Q structure corresponding to the highest-density Skyrmion lattice. Using inelastic neutron scattering (INS) and advanced spin dynamics simulations, we demonstrate a clear distinction in the excitation spectra between the single-Q and triple-Q phases of Co1/3TaS2 and, more generally, a triangular lattice. First, we refined the spin Hamiltonian by fitting the excitation spectra measured in its paramagnetic phase, allowing us to develop an unbiased model independent of magnetic ordering. Second, we observed that the two magnetically ordered phases in Co1/3TaS2 exhibit markedly different behaviors in their long-wavelength Goldstone modes. Our spin model, derived from the paramagnetic phase, confirms that these behaviors originate from the single-Q and triple-Q nature of the respective ordered phases, providing unequivocal evidence of the single-Q to triple-Q phase transition in Co1/3TaS2. Importantly, we propose that the observed contrast in the long-wavelength spin dynamics between the single-Q and triple-Q orderings is universal, offering a potentially unique way to distinguish a generic triple-Q ordering on a triangular lattice from its multi-domain single-Q counterparts. We describe its applicability with examples of similar hexagonal systems forming potential triple-Q orderings. (For the full abstract, please refer to the manuscript), Comment: 14 pages, 6 figures excluding Appendices

Large Language Models (LLMs) typically generate outputs token by token using a fixed compute budget, leading to inefficient resource utilization. To address this shortcoming, recent advancements in mixture of expert (MoE) models, speculative decoding, and early exit strategies leverage the insight that computational demands can vary significantly based on the complexity and nature of the input. However, identifying optimal routing patterns for dynamic execution remains an open challenge, limiting the full potential of these adaptive methods. To address this need, we study adaptive computation in LLMs more systematically. We propose a novel framework that integrates smaller auxiliary modules within each Feed-Forward Network layer of the LLM. This design enables dynamic routing of tokens based on task complexity: tokens can be processed by either the small or big modules at each layer, or even bypass certain layers entirely. This allows us to introduce a novel notion of a token's difficulty, defined by its potential to benefit from additional computational resources. Importantly, by employing oracles to identify optimal patterns of adaptive computations, we gain valuable insights into the internal workings of LLMs and the routing processes in a simplified heterogeneous MoE setup. We show that trained routers operate differently from oracles and often yield suboptimal solutions. Notably, activating a large module in just one layer outperforms models that use large modules across all layers, underscoring the gap between practical implementations of routing in MoE models and theoretical optima for adaptive computation.

In this paper, we establish for the first time the oracle complexity of a Riemannian inexact augmented Lagrangian (RiAL) method with the classical dual update for solving a class of Riemannian nonsmooth composite problems. By using the Riemannian gradient descent method with a specified stopping criterion for solving the inner subproblem, we show that the RiAL method can find an $\varepsilon$-stationary point of the considered problem with $\mathcal{O}(\varepsilon^{-3})$ calls to the first-order oracle. This achieves the best oracle complexity known to date. Numerical results demonstrate that the use of the classical dual stepsize is crucial to the high efficiency of the RiAL method., Comment: 16 pages, 3 tables

Correlations between galaxies and their supermassive black holes (SMBHs) have been observed, but the causal mechanisms remained unclear. The emerging field of causal inference now enables examining these relationships using observational data. This study, using advanced causal discovery techniques and a state-of-the-art dataset, reveals a causal link between galaxy properties and SMBH masses. In elliptical galaxies, bulge properties influence SMBH growth, while in spiral galaxies, SMBHs affect host galaxy properties, potentially through feedback in gas-rich environments. For spiral galaxies, SMBHs progressively quench star formation, whereas in elliptical galaxies, quenching is complete, and the causal connection has reversed. These findings support theoretical models of active galactic nuclei feedback regulating galaxy evolution and suggest further exploration of causal links in astrophysical and cosmological scaling relations., Comment: 34 pages, 20 figures, submitted for peer-review

Large-scale vision-language models like CLIP have demonstrated impressive open-vocabulary capabilities for image-level tasks, excelling in recognizing what objects are present. However, they struggle with pixel-level recognition tasks like semantic segmentation, which additionally require understanding where the objects are located. In this work, we propose a novel method, PixelCLIP, to adapt the CLIP image encoder for pixel-level understanding by guiding the model on where, which is achieved using unlabeled images and masks generated from vision foundation models such as SAM and DINO. To address the challenges of leveraging masks without semantic labels, we devise an online clustering algorithm using learnable class names to acquire general semantic concepts. PixelCLIP shows significant performance improvements over CLIP and competitive results compared to caption-supervised methods in open-vocabulary semantic segmentation. Project page is available at https://cvlab-kaist.github.io/PixelCLIP, Comment: To appear at NeurIPS 2024. Project page is available at https://cvlab-kaist.github.io/PixelCLIP

Spins in solid systems can inherently serve as qubits for quantum simulation or quantum information processing. Spin qubits are usually prone to environmental magnetic field fluctuations; however, a spin qubit encoded in a decoherence-free-subspace (DFS) can be protected from certain degrees of environmental noise depending on the specific structure of the DFS. Here, we derive the "true" DFS from an antiferromagnetic Heisenberg spin-1/2 trimer, which protects the qubit states against both short- and long-wavelength magnetic field fluctuations. We define the spin trimer with three electrons confined in a gate-defined GaAs double quantum dot (DQD) where we exploit Wigner-molecularization in one of the quantum dots. We first utilize the trimer for dynamic nuclear polarization (DNP), which generates a sizable magnetic field difference, $\Delta B_\mathrm{z}$, within the DQD. We show that large $\Delta B_\mathrm{z}$ significantly alters the eigenspectrum of the trimer and results in the "true" DFS in the DQD. Real-time Bayesian estimation of the DFS energy gap explicitly demonstrates protection of the DFS against short-wavelength magnetic field fluctuations in addition to long-wavelength ones. Our findings pave the way toward compact DFS structures for exchange-coupled quantum dot spin chains, the internal structure of which can be coherently controlled completely decoupled from environmental magnetic fields.

Recently, there has been growing interest in minimax problems on Riemannian manifolds due to their wide applications in machine learning and signal processing. Although many algorithms have been developed for minimax problems in the Euclidean setting, there are relatively few works studying minimax problems on manifolds. In this paper, we develop a flexible Riemannian alternating descent ascent (RADA) algorithmic framework for solving nonconvex-linear minimax problems on Riemannian manifolds. Within this framework, we propose two easy-to-implement yet efficient algorithms that alternately perform one or multiple projected/Riemannian gradient descent steps and a proximal gradient ascent step at each iteration. We show that the proposed RADA algorithmic framework can find both an $\varepsilon$-Riemannian-game-stationary point and an $\varepsilon$-Riemannian-optimization-stationary point of the considered problem within $\mathcal{O}(\varepsilon^{-3})$ iterations, achieving the best-known iteration complexity. We also reveal intriguing similarities and differences between the algorithms developed within our proposed framework and existing algorithms, which provide important insights into why the former outperform the latter. Lastly, we report numerical results on sparse principal component analysis (PCA), fair PCA, and sparse spectral clustering to demonstrate the superior performance of the proposed algorithms., Comment: 39 pages, 5 tables

Clinical experts employ diverse philosophies and strategies in patient care, influenced by regional patient populations. However, existing medical artificial intelligence (AI) models are often trained on data distributions that disproportionately reflect highly prevalent patterns, reinforcing biases and overlooking the diverse expertise of clinicians. To overcome this limitation, we introduce the Mixture of Multicenter Experts (MoME) approach. This method strategically integrates specialized expertise from diverse clinical strategies, enhancing the AI model's ability to generalize and adapt across multiple medical centers. The MoME-based multimodal target volume delineation model, trained with few-shot samples including images and clinical notes from each medical center, outperformed baseline methods in prostate cancer radiotherapy target delineation. The advantages of MoME were most pronounced when data characteristics varied across centers or when data availability was limited, demonstrating its potential for broader clinical applications. Therefore, the MoME framework enables the deployment of AI-based target volume delineation models in resource-constrained medical facilities by adapting to specific preferences of each medical center only using a few sample data, without the need for data sharing between institutions. Expanding the number of multicenter experts within the MoME framework will significantly enhance the generalizability, while also improving the usability and adaptability of clinical AI applications in the field of precision radiation oncology., Comment: 39 pages

This paper introduces a speech enhancement solution tailored for true wireless stereo (TWS) earbuds on-device usage. The solution was specifically designed to support conversations in noisy environments, with active noise cancellation (ANC) activated. The primary challenges for speech enhancement models in this context arise from computational complexity that limits on-device usage and latency that must be less than 3 ms to preserve a live conversation. To address these issues, we evaluated several crucial design elements, including the network architecture and domain, design of loss functions, pruning method, and hardware-specific optimization. Consequently, we demonstrated substantial improvements in speech enhancement quality compared with that in baseline models, while simultaneously reducing the computational complexity and algorithmic latency., Comment: Accepted by Interspeech 2024

Intra-body nanonetworks hold promise for advancing targeted drug delivery (TDD) systems through molecular communications (MC). In the baseline MC-TDD system, drug-loaded nanomachines (DgNs) are positioned near the infected tissues to deliver drug molecules directly. To mitigate the decline in drug delivery efficiency caused by diffusion, we propose an enhanced MC-TDD system with a relay network. This network employs a novel localization-enabled relaying mechanism, where a nano-controller broadcasts a localization signal. DgNs then measure the received signal strength against thresholds to determine their clusters relative to the infected tissue. Additionally, our study considers the effect of multiple absorbing DgNs on the channel impulse response (CIR), a factor overlooked in previous works. Our approach improves drug delivery efficiency by $17\%$ compared to the baseline system. Importantly, we find that optimizing CIR is crucial for enhancing drug delivery efficiency. These findings pave the way for further research into optimizing CIR-based relay selection, as well as investigating the impact of factors such as drug molecule lifespan, obstruction probabilities, and flow dynamics., Comment: Accepted at ACM NanoCom October 2024

Existing causal inference (CI) models are limited to primarily handling low-dimensional confounders and singleton actions. We propose an autoregressive (AR) CI framework capable of handling complex confounders and sequential actions common in modern applications. We accomplish this by {\em sequencification}, transforming data from an underlying causal diagram into a sequence of tokens. This approach not only enables training with data generated from any DAG but also extends existing CI capabilities to accommodate estimating several statistical quantities using a {\em single} model. We can directly predict interventional probabilities, simplifying inference and enhancing outcome prediction accuracy. We demonstrate that an AR model adapted for CI is efficient and effective in various complex applications such as navigating mazes, playing chess endgames, and evaluating the impact of certain keywords on paper acceptance rates.

This paper introduces the hypothetical track-length fitting algorithm, a novel method for measuring the kinetic energies of ionizing particles in liquid argon time projection chambers (LArTPCs). The algorithm finds the most probable offset in track length for a track-like object by comparing the measured ionization density as a function of position with a theoretical prediction of the energy loss as a function of the energy, including models of electron recombination and detector response. The algorithm can be used to measure the energies of particles that interact before they stop, such as charged pions that are absorbed by argon nuclei. The algorithm's energy measurement resolutions and fractional biases are presented as functions of particle kinetic energy and number of track hits using samples of stopping secondary charged pions in data collected by the ProtoDUNE-SP detector, and also in a detailed simulation. Additional studies describe impact of the dE/dx model on energy measurement performance. The method described in this paper to characterize the energy measurement performance can be repeated in any LArTPC experiment using stopping secondary charged pions.

Purpose: Diffusion weighted MRI (dMRI) and its models of neural structure provide insight into human brain organization and variations in white matter. A recent study by McMaster, et al. showed that complex graph measures of the connectome, the graphical representation of a tractogram, vary with spatial sampling changes, but biases introduced by anisotropic voxels in the process have not been well characterized. This study uses microstructural measures (fractional anisotropy and mean diffusivity) and white matter bundle properties (bundle volume, length, and surface area) to further understand the effect of anisotropic voxels on microstructure and tractography. Methods: The statistical significance of the selected measures derived from dMRI data were assessed by comparing three white matter bundles at different spatial resolutions with 44 subjects from the Human Connectome Project Young Adult dataset scan/rescan data using the Wilcoxon Signed Rank test. The original isotropic resolution (1.25 mm isotropic) was explored with six anisotropic resolutions with 0.25 mm incremental steps in the z dimension. Then, all generated resolutions were upsampled to 1.25 mm isotropic and 1 mm isotropic. Results: There were statistically significant differences between at least one microstructural and one bundle measure at every resolution (p less than or equal to 0.05, corrected for multiple comparisons). Cohen's d coefficient evaluated the effect size of anisotropic voxels on microstructure and tractography. Conclusion: Fractional anisotropy and mean diffusivity cannot be recovered with basic up sampling from low quality data with gold standard data. However, the bundle measures from tractogram become more repeatable when voxels are resampled to 1 mm isotropic.

In this paper, we present an effective data augmentation framework leveraging the Large Language Model (LLM) and Diffusion Model (DM) to tackle the challenges inherent in data-scarce scenarios. Recently, DMs have opened up the possibility of generating synthetic images to complement a few training images. However, increasing the diversity of synthetic images also raises the risk of generating samples outside the target distribution. Our approach addresses this issue by embedding novel semantic information into text prompts via LLM and utilizing real images as visual prompts, thus generating semantically rich images. To ensure that the generated images remain within the target distribution, we dynamically adjust the guidance weight based on each image's CLIPScore to control the diversity. Experimental results show that our method produces synthetic images with enhanced diversity while maintaining adherence to the target distribution. Consequently, our approach proves to be more efficient in the few-shot setting on several benchmarks. Our code is available at https://github.com/kkyuhun94/dalda ., Comment: Accepted to ECCV Synthetic Data for Computer Vision Workshop (Oral)

Understanding the origins of massive black hole seeds and their co-evolution with their host galaxy requires studying intermediate-mass black holes (IMBHs) and estimating their mass. However, measuring the mass of these IMBHs is challenging due to the high spatial resolution requirement. A spectrophotometric reverberation monitoring is performed for a low-luminosity Seyfert 1 galaxy NGC 4395 to measure the size of the broad line region (BLR) and black hole mass. The data were collected using the 1.3-m Devasthal fast optical telescope (DFOT) and 3.6-m Devasthal optical telescope (DOT) at ARIES, Nainital, over two consecutive days in March 2022. The analysis revealed strong emission lines in the spectra and light curves of merged 5100{\AA} spectroscopic continuum flux ($f_{\mathrm{5100}}$) with photometric continuum V-band and H$\alpha$, with fractional variabilities of 6.38\% and 6.31\% respectively. In comparison to several previous studies with lag estimation $<$ 90 minutes, our calculated H$\alpha$ lag supersedes by $125.0^{+6.2}_{-6.1}$ minutes using ICCF and {\small JAVELIN} methods. The velocity dispersion ($\sigma_{\mathrm{line}}$) of the broad line clouds is measured to be $544.7^{+22.4}_{-25.1}$ km s$^{-1}$, yielding a black hole mass of $\sim$ $2.2^{+0.2}_{-0.2}\times 10^{4}M_{\mathrm{\odot}}$ and an Eddington ratio of 0.06., Comment: 17 pages, 8 figures, accepted for publication in ApJ

We introduce a learning-guided motion planning framework that provides initial seed trajectories using a diffusion model for trajectory optimization. Given a workspace, our method approximates the configuration space (C-space) obstacles through a key-configuration representation that consists of a sparse set of task-related key configurations, and uses this as an input to the diffusion model. The diffusion model integrates regularization terms that encourage collision avoidance and smooth trajectories during training, and trajectory optimization refines the generated seed trajectories to further correct any colliding segments. Our experimental results demonstrate that using high-quality trajectory priors, learned through our C-space-grounded diffusion model, enables efficient generation of collision-free trajectories in narrow-passage environments, outperforming prior learning- and planning-based baselines. Videos and additional materials can be found on the project page: https://kiwi-sherbet.github.io/PRESTO., Comment: Submitted to ICRA 2025

Machine learning is attracting surging interest across nearly all scientific areas by enabling the analysis of large datasets and the extraction of scientific information from incomplete data. Data-driven science is rapidly growing, especially in X-ray methodologies, where advanced light sources and detection technologies accumulate vast amounts of data that exceed meticulous human inspection capabilities. Despite the increasing demands, the full application of machine learning has been hindered by the need for data-specific optimizations. In this study, we introduce a new deep-learning-based phase retrieval method for imperfect diffraction data. This method provides robust phase retrieval for simulated data and performs well on weak-signal single-pulse diffraction data from X-ray free-electron lasers. Moreover, the method significantly reduces data processing time, facilitating real-time image reconstructions that are crucial for high-repetition-rate data acquisition. Thus, this approach offers a reliable solution to the phase problem and is expected to be widely adopted across various research areas.

An upgrade of the COSINE-100 experiment, COSINE-100U, has been prepared for installation at Yemilab, a new underground laboratory in Korea, following 6.4 years of operation at the Yangyang Underground Laboratory. The COSINE-100 experiment aimed to investigate the annual modulation signals reported by the DAMA/LIBRA but observed a null result, revealing a more than 3$\sigma$ discrepancy. COSINE-100U seeks to explore new parameter spaces for dark matter detection using NaI(Tl) detectors. All eight NaI(Tl) crystals, with a total mass of 99.1 kg, have been upgraded to improve light collection efficiency, significantly enhancing dark matter detection sensitivity. This paper describes the detector upgrades, performance improvements, and the enhanced sensitivity to low-mass dark matter detection in the COSINE-100U experiment., Comment: 14 pages, 17 figures

Optical microscopy has long been the standard method for diagnosis in cytopathology. Whole slide scanners can image and digitize large sample areas automatically, but are slow, expensive and therefore not widely available. Clinical diagnosis of cytology specimens is especially challenging since these samples are both spread over large areas and thick, which requires 3D capture. Here, we introduce a new parallelized microscope for scanning thick specimens across extremely wide fields-of-view (54x72 mm^2) at 1.2 and 0.6 {\mu}m resolutions, accompanied by machine learning software to rapidly assess these 16 gigapixel scans. This Multi-Camera Array Scanner (MCAS) comprises 48 micro-cameras closely arranged to simultaneously image different areas. By capturing 624 megapixels per snapshot, the MCAS is significantly faster than most conventional whole slide scanners. We used this system to digitize entire cytology samples (scanning three entire slides in 3D in just several minutes) and demonstrate two machine learning techniques to assist pathologists: first, an adenocarcinoma detection model in lung specimens (0.73 recall); second, a slide-level classification model of lung smears (0.969 AUC).

We compare two amplification schemes, photon addition and then subtraction ($\hat{a}\hat{a}^\dagger$) and successive photon addition ($\hat{a}^\dagger{}^2$), applied to hybrid entangled states (HESs) and superpositions of coherent states (SCSs). We show that the amplification schemes' fidelity and gain for HESs are the same as those of coherent states. On the other hand, SCSs show quite nontrivial behaviors by the amplification schemes, depending on the amplitudes of coherent states, number of coherent-state components, and relative phases between the components. This implies that appropriate amplification schemes for SCSs should be chosen depending on the tasks and specific forms of the states. To investigate the quality of amplified states, we calculate the quantum Fisher information, a measure of quantum phase estimation. In terms of the quantum Fisher information, the $\hat{a}\hat{a}^\dagger$ scheme tends to show better performance for relatively small amplitudes while the $\hat{a}^\dagger{}^2$ scheme is better in larger amplitude regime. The performance of the two schemes becomes indistinguishable as the amplitude grows sufficiently large., Comment: 14 pages, 6 figures