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

Liang Gao

Liang Gao contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
astro-ph.GAastro-ph.COArtificial IntelligenceComputer Visioneess.ASMultimediaphysics.comp-phSound

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
10works
0followers
8topics
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

10 published item(s)

preprint2022arXiv

A New Knowledge Distillation Network for Incremental Few-Shot Surface Defect Detection

Surface defect detection is one of the most essential processes for industrial quality inspection. Deep learning-based surface defect detection methods have shown great potential. However, the well-performed models usually require large training data and can only detect defects that appeared in the training stage. When facing incremental few-shot data, defect detection models inevitably suffer from catastrophic forgetting and misclassification problem. To solve these problems, this paper proposes a new knowledge distillation network, called Dual Knowledge Align Network (DKAN). The proposed DKAN method follows a pretraining-finetuning transfer learning paradigm and a knowledge distillation framework is designed for fine-tuning. Specifically, an Incremental RCNN is proposed to achieve decoupled stable feature representation of different categories. Under this framework, a Feature Knowledge Align (FKA) loss is designed between class-agnostic feature maps to deal with catastrophic forgetting problems, and a Logit Knowledge Align (LKA) loss is deployed between logit distributions to tackle misclassification problems. Experiments have been conducted on the incremental Few-shot NEU-DET dataset and results show that DKAN outperforms other methods on various few-shot scenes, up to 6.65% on the mean Average Precision metric, which proves the effectiveness of the proposed method.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

About One-point Statistics of the Ratio of Two Fourier-transformed Cosmic Fields and an Application

The Fourier transformation is an effective and efficient operation of Gaussianization at the one-point level. Using a set of N-body simulation data, we verified that the one-point distribution functions of the dark matter momentum divergence and density fields closely follow complex Gaussian distributions. The one-point distribution function of the quotient of two complex Gaussian variables is introduced and studied. Statistical theories are then applied to model one-point statistics about the growth of individual Fourier mode of the dark matter density field, which can be obtained by the ratio of two Fourier transformed cosmic fields. Our simulation results proved that the models based on the Gaussian approximation are impressively accurate, and our analysis revealed many interesting aspects about the growth of dark matter's density fluctuation in Fourier space.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

The impact of filaments on dwarf galaxy properties in the Auriga simulations

With a hydrodynamical simulation using a simple galaxy formation model without taking into account feedback, our previous work has shown that dense and massive filaments at high redshift can provide potential wells to trap and compress gas, and hence affect galaxy formation in their resident low-mass haloes. In this paper, we make use of the Auriga simulations, a suite of high-resolution zoom-in hydrodynamical simulations of Milky Way-like galaxies, to study whether the conclusion still holds in the simulations with a sophisticated galaxy formation model. In agreement with the results of our previous work, we find that, comparing to their counterparts with similar halo masses in field, dwarf galaxies residing in filaments tend to have higher baryonic and stellar fractions. At the fixed parent halo mass, the filament dwarfs tend to have slightly higher star formation rates than those of field ones. But overall we do not find a clear difference in galaxy g - r colours between the filament and field populations. We also show that at high redshifts, the gas components in dwarf galaxies tend to have their spins aligned with the filaments in which they reside. Our results support a picture in which massive filaments at high redshift assist gas accretion and enhance star formation in their resident dwarf sized dark matter haloes.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Constraining the inner density slope of massive galaxy clusters

We determine the inner density profiles of massive galaxy clusters (M$_{200}$ > $5 \times 10^{14}$ M$_{\odot}$) in the Cluster-EAGLE (C-EAGLE) hydrodynamic simulations, and investigate whether the dark matter density profiles can be correctly estimated from a combination of mock stellar kinematical and gravitational lensing data. From fitting mock stellar kinematics and lensing data generated from the simulations, we find that the inner density slopes of both the total and the dark matter mass distributions can be inferred reasonably well. We compare the density slopes of C-EAGLE clusters with those derived by Newman et al. for 7 massive galaxy clusters in the local Universe. We find that the asymptotic best-fit inner slopes of "generalized" NFW (gNFW) profiles, $γ_{\rm gNFW}$, of the dark matter haloes of the C-EAGLE clusters are significantly steeper than those inferred by Newman et al. However, the mean mass-weighted dark matter density slopes of the simulated clusters are in good agreement with the Newman et al. estimates. We also find that the estimate of $γ_{\rm gNFW}$ is very sensitive to the constraints from weak lensing measurements in the outer parts of the cluster and a bias can lead to an underestimate of $γ_{\rm gNFW}$.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Multi-Representation Knowledge Distillation For Audio Classification

As an important component of multimedia analysis tasks, audio classification aims to discriminate between different audio signal types and has received intensive attention due to its wide applications. Generally speaking, the raw signal can be transformed into various representations (such as Short Time Fourier Transform and Mel Frequency Cepstral Coefficients), and information implied in different representations can be complementary. Ensembling the models trained on different representations can greatly boost the classification performance, however, making inference using a large number of models is cumbersome and computationally expensive. In this paper, we propose a novel end-to-end collaborative learning framework for the audio classification task. The framework takes multiple representations as the input to train the models in parallel. The complementary information provided by different representations is shared by knowledge distillation. Consequently, the performance of each model can be significantly promoted without increasing the computational overhead in the inference stage. Extensive experimental results demonstrate that the proposed approach can improve the classification performance and achieve state-of-the-art results on both acoustic scene classification tasks and general audio tagging tasks.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Simulating Kilonovae in the ΛCDM Universe

Kilonovae are optical flashes produced in the aftermath of neutron star-neutron star mergers (NNMs) or neutron star-black hole mergers (NBMs). In this work, we use the Millennium Simulation, combined with a semi-analytic galaxy formation model--GABE (Galaxy Assembly with Binary Evolution) which adopts binary stellar population synthesis models, to explore the cosmic event rate of kilonovae, and the properties of their host galaxies in a cosmological context. We find that model with supernova kick velocity of 0 km/s fits the observation best, in agreement with the exception of some formation channels of binary neutron star. This indicates that NNMs prefer to originate from binary systems with low kick velocities. With V$_{\rm kick}$=0 km/s, the cosmic event rate of NNMs and NBMs at z=0 are 283 Gpc$^{-3}$yr$^{-1}$ and 91 Gpc$^{-3}$yr$^{-1}$, respectively, marginally consistent with the constraint from LVC GWTC-1. For Milky Way-mass galaxies, we predict the NNM rate is $25.7^{+59.6}_{-7.1}$ Myr$^{-1}$, which is also in good agreement with the observed properties of binary neutron stars in the Milky Way. Taking all the NNMs into account in the history of Milky Way-mass galaxies, we find that the averaged r-process elements yield with A>79 in a NNM and NBM event should be 0.01 M$_{\odot}$ to be consistent with observation. We conclude that NGC 4993, the host galaxy of GW170817, is a typical host galaxy for NNMs. However, generally NNMs and NBMs tend to reside in young, blue, star-forming, late-type galaxies, with stellar mass and gaseous metallicity distribution peaking at $10^{10.65}$ M$_{\odot}$ and 8.72-8.85, respectively. By studying kilonovae host galaxies in the cosmological background, it is promising to constrain model details better when we have more events in the forthcoming future. (abridged)

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

The effect of hydrodynamics alone on the subhalo population in a $Λ$CDM rich cluster sized dark matter halo

We perform a set of non-radiative hydro-dynamical (NHD) simulations of a rich cluster sized dark matter halo from the Phoenix project with 3 different numerical resolutions, to investigate the effect of hydrodynamics alone on the subhalo population in the halo. Compared to dark matter only (DMO) simulations of the same halo, subhaloes are less abundant for relatively massive subhaloes ($M_{sub} > 2.5 \times 10^9h^{-1}M_{\odot}$, or $V_{max} > 70 kms^{-1}$) but more abundant for less massive subhaloes in the NHD simulations. This results in different shapes in the subhalo mass/$V_{max}$ function in two different sets of simulations. At given subhalo mass, the subhaloes less massive than $10^{10} h^{-1}M_{\odot}$ have larger $V_{max}$ in the NHD than DMO simulations, while $V_{max}$ is similar for the subhaloes more massive than the mass value. This is mainly because the progenitors of present day low mass subhaloes have larger concentration parameters in the NHD than DMO simulations. The survival number fraction of the accreted low mass progenitors of the main halo at redshift 2 is about 50 percent higher in the NHD than DMO simulations.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

The temperature of IGM at high redshifts: shock heating and high mach problem

The thermal history of cosmic gas in the Dark Ages remains largely unknown. It is important to quantify the impact of relevant physics on the IGM temperature between $z=10$ and $z \sim 30$, in order to interpret recent and oncoming observations, including results reported by EDGES. We revisit the gas heating due to structure formation shocks in this era, using a set of fixed grid cosmological hydrodynamical simulations performed by three different codes. In all our simulations, the cosmic gas is predicted to be in multiphase state since $z>30$. The gas surrounding high density peaks gradually develops a relation more sharp than $T \propto ρ^{2/3}$, approximately $T \propto ρ^{2}$, from $z=30$ to $z=11$, might due to shock heating. Meanwhile, the gas in void region tends to have a large local mach number, and their thermal state varies significantly from code to code. In the redshift range $11-20$, the mass fraction of gas shock heated above the CMB temperature in our simulations is larger than previous semi-analytical results by a factor of 2 to 8. At $z=15$, the fraction varies from $\sim 19\%$ to $52 \%$ among different codes. Between $z=11$ and $z=20$, the gas temperature $<1/T_{\rm{K}}>_M^{-1}$ is predicted to be $\sim 10-20$ K by two codes, much higher than the adiabatic cooling model and some previous works. However, in our simulations performed by RAMSES, $<1/T_{\rm{K}}>_M^{-1}$ is predicted to be even below the temperature required to explain result of the EDGES. Given the fact that different codes give different predictions, currently, it seems a challenge to make solid prediction on the temperature of gas at $z \sim 17$ in simulations.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Universal structure of dark matter haloes over a mass range of 20 orders of magnitude

Cosmological models in which dark matter consists of cold elementary particles predict that the dark halo population should extend to masses many orders of magnitude below those at which galaxies can form. Here we report a cosmological simulation of the formation of present-day haloes over the full range of observed halo masses (20 orders of magnitude) when dark matter is assumed to be in the form of weakly interacting massive particles of mass approximately 100 gigaelectronvolts. The simulation has a full dynamic range of 30 orders of magnitude in mass and resolves the internal structure of hundreds of Earth-mass haloes in as much detail as it does for hundreds of rich galaxy clusters. We find that halo density profiles are universal over the entire mass range and are well described by simple two-parameter fitting formulae. Halo mass and concentration are tightly related in a way that depends on cosmology and on the nature of the dark matter. For a fixed mass, the concentration is independent of the local environment for haloes less massive than those of typical galaxies. Haloes over the mass range of 10^3 to 10^11 solar masses contribute about equally (per logarithmic interval) to the luminosity produced by dark matter annihilation, which we find to be smaller than all previous estimates by factors ranging up to one thousand.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

HIKER: a halo-finding method based on kernel-shift algorithm

We introduce a new halo/subhalo finder, HIKER (a Halo fInder based on KERnel-shift algorithm), which takes advantage of a machine learning method -- the mean-shift algorithm combined with the Plummer kernel function, to effectively locate density peaks corresponding to halos/subhalos in density field. Based on these density peaks, dark matter halos are identified as spherical overdensity structures, and subhalos are bound substructures with boundaries at their tidal radius. By testing HIKER code with mock halos, we show that HIKER performs excellently in recovering input halo properties. Especially, HIKER has higher accuracy in locating halo/subhalo centres than most halo finders. With cosmological simulations, we further show that HIKER reproduces the abundance of dark matter halos and subhalos quite accurately, and the HIKER halo/subhalo mass functions and $V_{max}$ functions are in good agreement with two widely used halo finders, SUBFIND and AHF.

0 citations0 reviewsNo code linkedOpen access