BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Yin Li

Yin Li contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Computer Visionastro-ph.COMachine Learningastro-ph.GAastro-ph.IMArtificial Intelligenceeess.IVmath.SGastro-ph.EPCryptography and SecurityDatabasesInformation RetrievalInformation Theorymath.AGmath.ITmath.RTphysics.flu-dyn

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
24works
0followers
17topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

24 published item(s)

preprint2026arXiv

EgoIntrospect: An Egocentric Dataset and Benchmark for User-Centric Internal State Reasoning

Despite extensive efforts on egocentric video datasets and benchmarks, understanding users' internal states, which is crucial for enabling seamless AI assistant experiences, remains largely overlooked. In this work, we introduce EgoIntrospect, the first egocentric dataset captured in user-driven scenarios with self-annotations that explicitly reveal users' interactive intentions with AI assistants. EgoIntrospect was collected using a cross-device setup, providing synchronized video, audio, gaze, motion, and physiological signals. It consists of 180 hours of recordings from 60 subjects, with an average recording duration of 3 hours per subject. Leveraging EgoIntrospect, we formalize a suite of tasks centered on user internal states, including affective experience, interactive intent, and cognitive memory. We further process the annotations to construct benchmarks that evaluate the ability of modern multimodal large language models to reason about users' internal states from egocentric observations. Experiments on our benchmark suggest that existing multimodal large language models struggle to effectively leverage multimodal signals to infer users' subjective internal states. The dataset and annotations will be made publicly available to advance research in egocentric vision and wearable AI assistants. Project page: https://ego-introspect.github.io/

0 citations0 reviewsNo code linkedOpen access
preprint2025arXiv

Decomposing LLM Self-Correction: The Accuracy-Correction Paradox and Error Depth Hypothesis

Large Language Models (LLMs) are widely believed to possess self-correction capabilities, yet recent studies suggest that intrinsic self-correction--where models correct their own outputs without external feedback--remains largely ineffective. In this work, we systematically decompose self-correction into three distinct sub-capabilities: error detection, error localization, and error correction. Through cross-model experiments on GSM8K-Complex (n=500 per model, 346 total errors) with three major LLMs, we uncover a striking Accuracy-Correction Paradox: weaker models (GPT-3.5, 66% accuracy) achieve 1.6x higher intrinsic correction rates than stronger models (DeepSeek, 94% accuracy)--26.8% vs 16.7%. We propose the Error Depth Hypothesis: stronger models make fewer but deeper errors that resist self-correction. Error detection rates vary dramatically across architectures (10% to 82%), yet detection capability does not predict correction success--Claude detects only 10% of errors but corrects 29% intrinsically. Surprisingly, providing error location hints hurts all models. Our findings challenge linear assumptions about model capability and self-improvement, with important implications for the design of self-refinement pipelines.

0 citations0 reviewsNo code linkedOpen access
preprint2025arXiv

Fukaya categories of hyperplane arrangements

To a simple polarized hyperplane arrangement (not necessarily cyclic) $\mathbb{V}$, one can associate a stopped Liouville manifold (equivalently, a Liouville sector) $\left(M(\mathbb{V}),ξ\right)$, where $M(\mathbb{V})$ is the complement of finitely many hyperplanes in $\mathbb{C}^d$, obtained as the complexifications of the real hyperplanes in $\mathbb{V}$. The Liouville structure on $M(\mathbb{V})$ comes from a very affine embedding, and the stop $ξ$ is determined by the polarization. In this article, we study the symplectic topology of $\left(M(\mathbb{V}),ξ\right)$. In particular, we prove that their partially wrapped Fukaya categories are generated by Lagrangian submanifolds associated to the bounded and feasible chambers of $\mathbb{V}$. A computation of the Fukaya $A_\infty$-algebra of these Lagrangians then enables us to identity these wrapped Fukaya categories with the $\mathbb{G}_m^d$-equivariant hypertoric convolution algebras $\widetilde{B}(\mathbb{V})$ associated to $\mathbb{V}$. This confirms a conjecture of Lauda-Licata-Manion (arXiv:2009.03981) and provides evidence for the general conjecture of Lekili-Segal (arXiv:2304.10969) on the equivariant Fukaya categories of symplectic manifolds with Hamiltonian torus actions.

0 citations0 reviewsNo code linkedOpen access
preprint2025arXiv

Persistence of unknottedness of clean Lagrangian intersections

Let $Q_0$ and $Q_1$ be two Lagrangian spheres in a $6$-dimensional symplectic manifold. Assume that $Q_0$ and $Q_1$ intersect cleanly along a circle that is unknotted in both $Q_0$ and $Q_1$. We prove that there is no nearby Hamiltonian isotopy of $Q_0$ and $Q_1$ to a pair of Lagrangian spheres meeting cleanly along a circle that is knotted in either component, answering a question of Smith. The proof is based on a classification of the spherical summands in the prime decomposition of an exact Lagrangian in the Stein neighborhood of the union $Q_0\cup Q_1$ and the deep result that lens space rational Dehn surgeries characterize the unknot.

0 citations0 reviewsNo code linkedOpen access
preprint2024arXiv

Particle clustering in turbulence: Prediction of spatial and statistical properties with deep learning

We investigate the utility of deep learning for modeling the clustering of particles that are aerodynamically coupled to turbulent fluids. Using a Lagrangian particle module within the Athena++ hydrodynamics code, we simulate the dynamics of particles in the Epstein drag regime within a periodic domain of isotropic forced hydrodynamic turbulence. This setup is an idealized model relevant to the collisional growth of micron to mm-sized dust particles in early stage planet formation. The simulation data are used to train a U-Net deep learning model to predict gridded three-dimensional representations of the particle density and velocity fields, given as input the corresponding fluid fields. The trained model qualitatively captures the filamentary structure of clustered particles in a highly non-linear regime. We assess model fidelity by calculating metrics of the density field (the radial distribution function) and of the velocity field (the relative velocity and the relative radial velocity between particles). Although trained only on the spatial fields, the model predicts these statistical quantities with errors that are typically <10%. Our results suggest that, given appropriately expanded training data, deep learning could complement direct numerical simulations in predicting particle clustering within turbulent flows.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

ActionFormer: Localizing Moments of Actions with Transformers

Self-attention based Transformer models have demonstrated impressive results for image classification and object detection, and more recently for video understanding. Inspired by this success, we investigate the application of Transformer networks for temporal action localization in videos. To this end, we present ActionFormer -- a simple yet powerful model to identify actions in time and recognize their categories in a single shot, without using action proposals or relying on pre-defined anchor windows. ActionFormer combines a multiscale feature representation with local self-attention, and uses a light-weighted decoder to classify every moment in time and estimate the corresponding action boundaries. We show that this orchestrated design results in major improvements upon prior works. Without bells and whistles, ActionFormer achieves 71.0% mAP at tIoU=0.5 on THUMOS14, outperforming the best prior model by 14.1 absolute percentage points. Further, ActionFormer demonstrates strong results on ActivityNet 1.3 (36.6% average mAP) and EPIC-Kitchens 100 (+13.5% average mAP over prior works). Our code is available at http://github.com/happyharrycn/actionformer_release.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Cosmology with one galaxy?

Galaxies can be characterized by many internal properties such as stellar mass, gas metallicity, and star-formation rate. We quantify the amount of cosmological and astrophysical information that the internal properties of individual galaxies and their host dark matter halos contain. We train neural networks using hundreds of thousands of galaxies from 2,000 state-of-the-art hydrodynamic simulations with different cosmologies and astrophysical models of the CAMELS project to perform likelihood-free inference on the value of the cosmological and astrophysical parameters. We find that knowing the internal properties of a single galaxy allow our models to infer the value of $Ω_{\rm m}$, at fixed $Ω_{\rm b}$, with a $\sim10\%$ precision, while no constraint can be placed on $σ_8$. Our results hold for any type of galaxy, central or satellite, massive or dwarf, at all considered redshifts, $z\leq3$, and they incorporate uncertainties in astrophysics as modeled in CAMELS. However, our models are not robust to changes in subgrid physics due to the large intrinsic differences the two considered models imprint on galaxy properties. We find that the stellar mass, stellar metallicity, and maximum circular velocity are among the most important galaxy properties to determine the value of $Ω_{\rm m}$. We believe that our results can be explained taking into account that changes in the value of $Ω_{\rm m}$, or potentially $Ω_{\rm b}/Ω_{\rm m}$, affect the dark matter content of galaxies. That effect leaves a distinct signature in galaxy properties to the one induced by galactic processes. Our results suggest that the low-dimensional manifold hosting galaxy properties provides a tight direct link between cosmology and astrophysics.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Egocentric Activity Recognition and Localization on a 3D Map

Given a video captured from a first person perspective and the environment context of where the video is recorded, can we recognize what the person is doing and identify where the action occurs in the 3D space? We address this challenging problem of jointly recognizing and localizing actions of a mobile user on a known 3D map from egocentric videos. To this end, we propose a novel deep probabilistic model. Our model takes the inputs of a Hierarchical Volumetric Representation (HVR) of the 3D environment and an egocentric video, infers the 3D action location as a latent variable, and recognizes the action based on the video and contextual cues surrounding its potential locations. To evaluate our model, we conduct extensive experiments on the subset of Ego4D dataset, in which both human naturalistic actions and photo-realistic 3D environment reconstructions are captured. Our method demonstrates strong results on both action recognition and 3D action localization across seen and unseen environments. We believe our work points to an exciting research direction in the intersection of egocentric vision, and 3D scene understanding.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Event Neural Networks

Video data is often repetitive; for example, the contents of adjacent frames are usually strongly correlated. Such redundancy occurs at multiple levels of complexity, from low-level pixel values to textures and high-level semantics. We propose Event Neural Networks (EvNets), which leverage this redundancy to achieve considerable computation savings during video inference. A defining characteristic of EvNets is that each neuron has state variables that provide it with long-term memory, which allows low-cost, high-accuracy inference even in the presence of significant camera motion. We show that it is possible to transform a wide range of neural networks into EvNets without re-training. We demonstrate our method on state-of-the-art architectures for both high- and low-level visual processing, including pose recognition, object detection, optical flow, and image enhancement. We observe roughly an order-of-magnitude reduction in computational costs compared to conventional networks, with minimal reductions in model accuracy.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Physics to the Rescue: Deep Non-line-of-sight Reconstruction for High-speed Imaging

Computational approach to imaging around the corner, or non-line-of-sight (NLOS) imaging, is becoming a reality thanks to major advances in imaging hardware and reconstruction algorithms. A recent development towards practical NLOS imaging, Nam et al. demonstrated a high-speed non-confocal imaging system that operates at 5Hz, 100x faster than the prior art. This enormous gain in acquisition rate, however, necessitates numerous approximations in light transport, breaking many existing NLOS reconstruction methods that assume an idealized image formation model. To bridge the gap, we present a novel deep model that incorporates the complementary physics priors of wave propagation and volume rendering into a neural network for high-quality and robust NLOS reconstruction. This orchestrated design regularizes the solution space by relaxing the image formation model, resulting in a deep model that generalizes well on real captures despite being exclusively trained on synthetic data. Further, we devise a unified learning framework that enables our model to be flexibly trained using diverse supervision signals, including target intensity images or even raw NLOS transient measurements. Once trained, our model renders both intensity and depth images at inference time in a single forward pass, capable of processing more than 5 captures per second on a high-end GPU. Through extensive qualitative and quantitative experiments, we show that our method outperforms prior physics and learning based approaches on both synthetic and real measurements. We anticipate that our method along with the fast capturing system will accelerate future development of NLOS imaging for real world applications that require high-speed imaging.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Super-resolving Dark Matter Halos using Generative Deep Learning

Generative deep learning methods built upon Convolutional Neural Networks (CNNs) provide a great tool for predicting non-linear structure in cosmology. In this work we predict high resolution dark matter halos from large scale, low resolution dark matter only simulations. This is achieved by mapping lower resolution to higher resolution density fields of simulations sharing the same cosmology, initial conditions and box-sizes. To resolve structure down to a factor of 8 increase in mass resolution, we use a variation of U-Net with a conditional GAN, generating output that visually and statistically matches the high resolution target extremely well. This suggests that our method can be used to create high resolution density output over Gpc/h box-sizes from low resolution simulations with negligible computational effort.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

The ASTRID simulation: the evolution of Supermassive Black Holes

We present the evolution of black holes (BHs) and their relationship with their host galaxies in Astrid, a large-volume cosmological hydrodynamical simulation with box size 250 $h^{-1} \rm Mpc$ containing $2\times5500^3$ particles evolved to z=3. Astrid statistically models BH gas accretion and AGN feedback to their environments, applies a power-law distribution for BH seed mass $M_{\rm sd}$, uses a dynamical friction model for BH dynamics and executes a physical treatment of BH mergers. The BH population is broadly consistent with empirical constraints on the BH mass function, the bright end of the luminosity functions, and the time evolution of BH mass and accretion rate density. The BH mass and accretion exhibit a tight correlation with host stellar mass and star formation rate. We trace BHs seeded before z>10 down to z=3, finding that BHs carry virtually no imprint of the initial $M_{\rm sd}$ except those with the smallest $M_{\rm sd}$, where less than 50\% of them have doubled in mass. Gas accretion is the dominant channel for BH growth compared to BH mergers. With dynamical friction, Astrid predicts a significant delay for BH mergers after the first encounter of a BH pair, with a typical elapse time of about 200 Myrs. There are in total $4.5 \times 10^5$ BH mergers in Astrid at z>3, $\sim 10^3$ of which have X-ray detectable EM counterparts: a bright kpc scale dual AGN with $L_X>10^{43}$ erg/s. BHs with $M_{\rm BH} \sim 10^{7-8} M_{\odot}$ experience the most frequent mergers. Galaxies that host BH mergers are unbiased tracers of the overall $M_{\rm BH} - M_{*}$ relation. Massive ($>10^{11} M_{\odot}$) galaxies have a high occupation number (>10) of BHs, and hence host the majority of BH mergers.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Towards Non-Line-of-Sight Photography

Non-line-of-sight (NLOS) imaging is based on capturing the multi-bounce indirect reflections from the hidden objects. Active NLOS imaging systems rely on the capture of the time of flight of light through the scene, and have shown great promise for the accurate and robust reconstruction of hidden scenes without the need for specialized scene setups and prior assumptions. Despite that existing methods can reconstruct 3D geometries of the hidden scene with excellent depth resolution, accurately recovering object textures and appearance with high lateral resolution remains an challenging problem. In this work, we propose a new problem formulation, called NLOS photography, to specifically address this deficiency. Rather than performing an intermediate estimate of the 3D scene geometry, our method follows a data-driven approach and directly reconstructs 2D images of a NLOS scene that closely resemble the pictures taken with a conventional camera from the location of the relay wall. This formulation largely simplifies the challenging reconstruction problem by bypassing the explicit modeling of 3D geometry, and enables the learning of a deep model with a relatively small training dataset. The results are NLOS reconstructions of unprecedented lateral resolution and image quality.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Weak gravitational lensing shear measurement with FPFS: analytical mitigation of noise bias and selection bias

Dedicated &#39;Stage IV&#39; observatories will soon observe the entire extragalactic sky, to measure the &#39;cosmic shear&#39; distortion of galaxy shapes by weak gravitational lensing. To measure the apparent shapes of those galaxies, we present an improved version of the Fourier Power Function Shapelets (FPFS) shear measurement method. This now includes analytic corrections for sources of bias that plague all shape measurement algorithms: including noise bias (due to noise in nonlinear combinations of observable quantities) and selection bias (due to sheared galaxies being more or less likely to be detected). Crucially, these analytic solutions do not rely on calibration from external image simulations. For isolated galaxies, the small residual $\sim$$10^{-3}$ multiplicative bias and $\lesssim$$10^{-4}$ additive bias now meet science requirements for Stage IV experiments. FPFS also works accurately for faint galaxies and robustly against stellar contamination. Future work will focus on deblending overlapping galaxies. The code used for this paper can process $>$$1000$ galaxy images per CPU second and is available from https://github.com/mr-superonion/FPFS.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

NECOLA: Towards a Universal Field-level Cosmological Emulator

We train convolutional neural networks to correct the output of fast and approximate N-body simulations at the field level. Our model, Neural Enhanced COLA --NECOLA--, takes as input a snapshot generated by the computationally efficient COLA code and corrects the positions of the cold dark matter particles to match the results of full N-body Quijote simulations. We quantify the accuracy of the network using several summary statistics, and find that NECOLA can reproduce the results of the full N-body simulations with sub-percent accuracy down to $k\simeq1~h{\rm Mpc}^{-1}$. Furthermore, the model, that was trained on simulations with a fixed value of the cosmological parameters, is also able to correct the output of COLA simulations with different values of $Ω_{\rm m}$, $Ω_{\rm b}$, $h$, $n_s$, $σ_8$, $w$, and $M_ν$ with very high accuracy: the power spectrum and the cross-correlation coefficients are within $\simeq1\%$ down to $k=1~h{\rm Mpc}^{-1}$. Our results indicate that the correction to the power spectrum from fast/approximate simulations or field-level perturbation theory is rather universal. Our model represents a first step towards the development of a fast field-level emulator to sample not only primordial mode amplitudes and phases, but also the parameter space defined by the values of the cosmological parameters.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

The CAMELS Multifield Dataset: Learning the Universe&#39;s Fundamental Parameters with Artificial Intelligence

We present the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) Multifield Dataset, CMD, a collection of hundreds of thousands of 2D maps and 3D grids containing many different properties of cosmic gas, dark matter, and stars from 2,000 distinct simulated universes at several cosmic times. The 2D maps and 3D grids represent cosmic regions that span $\sim$100 million light years and have been generated from thousands of state-of-the-art hydrodynamic and gravity-only N-body simulations from the CAMELS project. Designed to train machine learning models, CMD is the largest dataset of its kind containing more than 70 Terabytes of data. In this paper we describe CMD in detail and outline a few of its applications. We focus our attention on one such task, parameter inference, formulating the problems we face as a challenge to the community. We release all data and provide further technical details at https://camels-multifield-dataset.readthedocs.io.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Attention Distillation for Learning Video Representations

We address the challenging problem of learning motion representations using deep models for video recognition. To this end, we make use of attention modules that learn to highlight regions in the video and aggregate features for recognition. Specifically, we propose to leverage output attention maps as a vehicle to transfer the learned representation from a motion (flow) network to an RGB network. We systematically study the design of attention modules, and develop a novel method for attention distillation. Our method is evaluated on major action benchmarks, and consistently improves the performance of the baseline RGB network by a significant margin. Moreover, we demonstrate that our attention maps can leverage motion cues in learning to identify the location of actions in video frames. We believe our method provides a step towards learning motion-aware representations in deep models. Our project page is available at https://aptx4869lm.github.io/AttentionDistillation/

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Comprehensive Image Captioning via Scene Graph Decomposition

We address the challenging problem of image captioning by revisiting the representation of image scene graph. At the core of our method lies the decomposition of a scene graph into a set of sub-graphs, with each sub-graph capturing a semantic component of the input image. We design a deep model to select important sub-graphs, and to decode each selected sub-graph into a single target sentence. By using sub-graphs, our model is able to attend to different components of the image. Our method thus accounts for accurate, diverse, grounded and controllable captioning at the same time. We present extensive experiments to demonstrate the benefits of our comprehensive captioning model. Our method establishes new state-of-the-art results in caption diversity, grounding, and controllability, and compares favourably to latest methods in caption quality. Our project website can be found at http://pages.cs.wisc.edu/~yiwuzhong/Sub-GC.html.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Constraining the Halo Mass of Damped Ly$α$ Absorption Systems (DLAs) at $z=2-3.5$ using the Quasar-CMB Lensing Cross-correlation

We study the cross correlation of damped Ly$α$ systems (DLAs) and their background quasars, using the most updated DLA catalog and the Planck 2018 CMB lensing convergence field. Our measurement suggests that the DLA bias $b_{\rm DLA}$ is smaller than $3.1$, corresponding to $\log(M/M_\odot h^{-1})\leq 12.3$ at a confidence of $90\%$. These constraints are broadly consistent with Alonso et al. (2018) and previous measurements by cross-correlation between DLAs and the Ly$α$ forest (e.g. Font-Ribera et al. 2012; Perez-Rafols et al. 2018). Further, our results demonstrate the potential of obtaining a more precise measurement of the halo mass of high-redshift sources using next generation CMB experiments with a higher angular resolution. The python-based codes and data products of our analysis are available at https://github.com/LittleLin1999/CMB-lensingxDLA.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Forecasting Human-Object Interaction: Joint Prediction of Motor Attention and Actions in First Person Video

We address the challenging task of anticipating human-object interaction in first person videos. Most existing methods ignore how the camera wearer interacts with the objects, or simply consider body motion as a separate modality. In contrast, we observe that the international hand movement reveals critical information about the future activity. Motivated by this, we adopt intentional hand movement as a future representation and propose a novel deep network that jointly models and predicts the egocentric hand motion, interaction hotspots and future action. Specifically, we consider the future hand motion as the motor attention, and model this attention using latent variables in our deep model. The predicted motor attention is further used to characterise the discriminative spatial-temporal visual features for predicting actions and interaction hotspots. We present extensive experiments demonstrating the benefit of the proposed joint model. Importantly, our model produces new state-of-the-art results for action anticipation on both EGTEA Gaze+ and the EPIC-Kitchens datasets. Our project page is available at https://aptx4869lm.github.io/ForecastingHOI/

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Gradients as Features for Deep Representation Learning

We address the challenging problem of deep representation learning--the efficient adaption of a pre-trained deep network to different tasks. Specifically, we propose to explore gradient-based features. These features are gradients of the model parameters with respect to a task-specific loss given an input sample. Our key innovation is the design of a linear model that incorporates both gradient and activation of the pre-trained network. We show that our model provides a local linear approximation to an underlying deep model, and discuss important theoretical insights. Moreover, we present an efficient algorithm for the training and inference of our model without computing the actual gradient. Our method is evaluated across a number of representation-learning tasks on several datasets and using different network architectures. Strong results are obtained in all settings, and are well-aligned with our theoretical insights.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Interpretable and Accurate Fine-grained Recognition via Region Grouping

We present an interpretable deep model for fine-grained visual recognition. At the core of our method lies the integration of region-based part discovery and attribution within a deep neural network. Our model is trained using image-level object labels, and provides an interpretation of its results via the segmentation of object parts and the identification of their contributions towards classification. To facilitate the learning of object parts without direct supervision, we explore a simple prior of the occurrence of object parts. We demonstrate that this prior, when combined with our region-based part discovery and attribution, leads to an interpretable model that remains highly accurate. Our model is evaluated on major fine-grained recognition datasets, including CUB-200, CelebA and iNaturalist. Our results compare favorably to state-of-the-art methods on classification tasks, and our method outperforms previous approaches on the localization of object parts.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Measurement of Void Bias Using Separate Universe Simulations

Cosmic voids are biased tracers of the large-scale structure of the universe. Separate universe simulations (SUS) enable accurate measurements of this biasing relation by implementing the peak-background split (PBS). In this work, we apply the SUS technique to measure the void bias parameters. We confirm that the PBS argument works well for underdense tracers. The response of the void size distribution depends on the void radius. For voids larger (smaller) than the size at the peak of the distribution, the void abundance responds negatively (positively) to a long wavelength mode. The linear bias from the SUS is in good agreement with the cross power spectrum measurement on large scales. Using the SUS, we have detected the quadratic void bias for the first time in simulations. We find that $ b_2 $ is negative when the magnitude of $ b_1 $ is small, and that it becomes positive and increases rapidly when $ |b_1| $ increases. We compare the results from voids identified in the halo density field with those from the dark matter distribution, and find that the results are qualitatively similar, but the biases generally shift to the larger voids sizes.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Obscure: Information-Theoretically Secure, Oblivious, and Verifiable Aggregation Queries on Secret-Shared Outsourced Data -- Full Version

Despite exciting progress on cryptography, secure and efficient query processing over outsourced data remains an open challenge. We develop a communication-efficient and information-theoretically secure system, entitled Obscure for aggregation queries with conjunctive or disjunctive predicates, using secret-sharing. Obscure is strongly secure (i.e., secure regardless of the computational-capabilities of an adversary) and prevents the network, as well as, the (adversarial) servers to learn the user&#39;s queries, results, or the database. In addition, Obscure provides additional security features, such as hiding access-patterns (i.e., hiding the identity of the tuple satisfying a query) and hiding query-patterns (i.e., hiding which two queries are identical). Also, Obscure does not require any communication between any two servers that store the secret-shared data before/during/after the query execution. Moreover, our techniques deal with the secret-shared data that is outsourced by a single or multiple database owners, as well as, allows a user, which may not be the database owner, to execute the query over secret-shared data. We further develop (non-mandatory) privacy-preserving result verification algorithms that detect malicious behaviors, and experimentally validate the efficiency of Obscure on large datasets, the size of which prior approaches of secret-sharing or multi-party computation systems have not scaled to.

0 citations0 reviewsNo code linkedOpen access