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 Ma

Liang Ma contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Computer Visiongr-qcMachine Learninghep-thNetworking and Internet ArchitectureRoboticsArtificial Intelligencecond-mat.mes-hallcond-mat.mtrl-scieess.IVHardware ArchitectureHuman-Computer InteractionMethodologyphysics.chem-ph

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
21works
0followers
14topics
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

21 published item(s)

preprint2026arXiv

Mamoda2.5: Enhancing Unified Multimodal Model with DiT-MoE

We present Mamoda2.5, a unified AR-Diffusion framework that seamlessly integrates multimodal understanding and generation within a single architecture. To efficiently enhance the model's generation capability, we equip the Diffusion Transformer backbone with a fine-grained Mixture-of-Experts (MoE) design (128 experts, Top-8 routing), yielding a 25B-parameter model that activates only 3B parameters, significantly reducing training costs while scaling up the model capacity. Mamoda2.5 achieves top-tier generation performance on VBench 2.0 and sets a new record in video editing quality, surpassing evaluated open-source models and matching the performance of current top-tier proprietary models, including the Kling O1 on OpenVE-Bench. Furthermore, we introduce a joint few-step distillation and reinforcement learning framework that compresses the 30-step editing model into a 4-step model and greatly accelerates model inference. Compared to open-source baselines, Mamoda2.5 achieves up to $95.9\times$ faster video editing inference. In real-world applications, Mamoda2.5 has been successfully deployed for content moderation and creative restoration tasks in advertising scenarios, achieving a 98% success rate in internal advertising video editing scenario.

0 citations0 reviewsNo code linkedOpen access
preprint2025arXiv

Deep Deterministic Nonlinear ICA via Total Correlation Minimization with Matrix-Based Entropy Functional

Blind source separation, particularly through independent component analysis (ICA), is widely utilized across various signal processing domains for disentangling underlying components from observed mixed signals, owing to its fully data-driven nature that minimizes reliance on prior assumptions. However, conventional ICA methods rely on an assumption of linear mixing, limiting their ability to capture complex nonlinear relationships and to maintain robustness in noisy environments. In this work, we present deep deterministic nonlinear independent component analysis (DDICA), a novel deep neural network-based framework designed to address these limitations. DDICA leverages a matrix-based entropy function to directly optimize the independence criterion via stochastic gradient descent, bypassing the need for variational approximations or adversarial schemes. This results in a streamlined training process and improved resilience to noise. We validated the effectiveness and generalizability of DDICA across a range of applications, including simulated signal mixtures, hyperspectral image unmixing, modeling of primary visual receptive fields, and resting-state functional magnetic resonance imaging (fMRI) data analysis. Experimental results demonstrate that DDICA effectively separates independent components with high accuracy across a range of applications. These findings suggest that DDICA offers a robust and versatile solution for blind source separation in diverse signal processing tasks.

0 citations0 reviewsNo code linkedOpen access
preprint2025arXiv

Quadratic Curvature Correction to 5D Myers-Perry Metric

We consider quadratic curvature perturbation to the Myers-Perry black hole in five dimensions at the linear level in the coupling constant. The solution can then be solved order by order in terms of two dimensionless angular momentum parameters up to an arbitrary order. We present the results up to tenth order. The perturbed solution allows us to obtain the higher-derivative correction to the black hole thermodynamics, which we find is in complete agreement with the Reall-Santos method.

0 citations0 reviewsNo code linkedOpen access
preprint2023arXiv

Dyonic Black Strings and the Charge Lattice in Salam-Sezgin model

We obtain a class of dyonic black string solutions in 6D Salam-Sezgin model. We then calculate various thermodynamic quantities associated with this solution. Interestingly, for the thermodynamic quantities to be well defined, the temperature is bounded from above. However, the mass and entropy can still grow without any upper bound, reaching infinity at the maximal temperature. The quantization condition obeyed by various charges is also analyzed. In particular, we find that the Dirac quantization condition selects one particular sign choice for the magnetic string charges.

0 citations0 reviewsNo code linkedOpen access
preprint2023arXiv

Quantum Corrections to Pair Production of Charged Black Holes in de Sitter Space

We compute Euclidean action of charged de Sitter black holes in four dimensional gravitational Euler-Heisenberg model. It turns out that the action of a general Euclidean dyonically charged black hole is still controlled by the total entropy contributed by the black hole outer horizon and the cosmological horizon. For smooth configurations, the Euclidean action can be interpreted as the black hole production rate in de Sitter space. We show that the 4-derivative couplings break the symmetry between the production rate of the purely electric black hole and that of the purely magnetic black hole. Although electromagnetic duality is no longer a symmetry, it induces a transformation on the 4-derivative couplings, mapping the physical quantities of a purely electric black hole to those of a purely magnetic black hole and vice versa. We also observe that under the same transformation, unitarity constraints on the 4-derivative couplings remain invariant.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

$α'$-corrections to Near Extremal Dyonic Strings and Weak Gravity Conjecture

We construct non-extremal dyonic string solutions in 6D minimal supergravity where the leading higher derivative corrections arise from either the type IIA string theory compactified on K3 or the heterotic string theory compactified on 4-torus. The thermodynamical quantities and Euclidean actions of the strings are computed. In the near extremal regime, we calculate the force felt by a probe fundamental string in the background of the macroscopic dyonic string with leading $α'$ corrections. We find that in both the IIA and heterotic setups, away from extremality, the attractive force overwhelms the repulsive force. However, close to extremality, the $α'$ corrections can reduce the attractive force in the isoentropic process, where the charges are fixed. This feature may be used as a new constraint for supergravity models with consistent quantum gravity embedding, in cases where the extremal limit coincides with the BPS limit and the higher derivative corrections do not affect the mass-to-charge ratio. By contrast, the $α'$ corrections can enhance the attractive force in the isothermal or isoenergetic processes.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

A new simple dynamic muscle fatigue model and its validation

Musculoskeletal disorder (MSD) is one of the major health problems in physical work especially in manual handling jobs. In several literatures, muscle fatigue is considered to be closely related to MSD, especially for muscle related disorders. In addition to many existing analysis techniques for muscle fatigue assessment and MSD risk analysis, in this paper, a new muscle fatigue model was proposed. The new proposed model reflects the influence of external load, workload history, and individual differences. This model is simple in mathematics and can be easily applied in realtime calculation, such as the application in realtime virtual work simulation and evaluation. The new model was mathematically validated with 24 existing static models by comparing the calculated METs, and qualitatively or quantitatively validated with 3 existing dynamic models. The proposed model shows high or moderate similarities in predicting the METs with all the 24 static models. Validation results with the three dynamic models were also promising. The main limitation of the model is that it still lacks experimental validation for more dynamic situations. Relevance to industry Muscle fatigue is one of the main reasons causing MSDs in industry, especially for physical work. Correct evaluation of muscle fatigue is necessary to determine work-rest regimens and reduce the risks of MSD.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Deep Transfer Learning with Graph Neural Network for Sensor-Based Human Activity Recognition

The sensor-based human activity recognition (HAR) in mobile application scenarios is often confronted with sensor modalities variation and annotated data deficiency. Given this observation, we devised a graph-inspired deep learning approach toward the sensor-based HAR tasks, which was further used to build a deep transfer learning model toward giving a tentative solution for these two challenging problems. Specifically, we present a multi-layer residual structure involved graph convolutional neural network (ResGCNN) toward the sensor-based HAR tasks, namely the HAR-ResGCNN approach. Experimental results on the PAMAP2 and mHealth data sets demonstrate that our ResGCNN is effective at capturing the characteristics of actions with comparable results compared to other sensor-based HAR models (with an average accuracy of 98.18% and 99.07%, respectively). More importantly, the deep transfer learning experiments using the ResGCNN model show excellent transferability and few-shot learning performance. The graph-based framework shows good meta-learning ability and is supposed to be a promising solution in sensor-based HAR tasks.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Few-Shot Class-Incremental Learning by Sampling Multi-Phase Tasks

New classes arise frequently in our ever-changing world, e.g., emerging topics in social media and new types of products in e-commerce. A model should recognize new classes and meanwhile maintain discriminability over old classes. Under severe circumstances, only limited novel instances are available to incrementally update the model. The task of recognizing few-shot new classes without forgetting old classes is called few-shot class-incremental learning (FSCIL). In this work, we propose a new paradigm for FSCIL based on meta-learning by LearnIng Multi-phase Incremental Tasks (LIMIT), which synthesizes fake FSCIL tasks from the base dataset. The data format of fake tasks is consistent with the `real' incremental tasks, and we can build a generalizable feature space for the unseen tasks through meta-learning. Besides, LIMIT also constructs a calibration module based on transformer, which calibrates the old class classifiers and new class prototypes into the same scale and fills in the semantic gap. The calibration module also adaptively contextualizes the instance-specific embedding with a set-to-set function. LIMIT efficiently adapts to new classes and meanwhile resists forgetting over old classes. Experiments on three benchmark datasets (CIFAR100, miniImageNet, and CUB200) and large-scale dataset, i.e., ImageNet ILSVRC2012 validate that LIMIT achieves state-of-the-art performance.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Forward Compatible Few-Shot Class-Incremental Learning

Novel classes frequently arise in our dynamically changing world, e.g., new users in the authentication system, and a machine learning model should recognize new classes without forgetting old ones. This scenario becomes more challenging when new class instances are insufficient, which is called few-shot class-incremental learning (FSCIL). Current methods handle incremental learning retrospectively by making the updated model similar to the old one. By contrast, we suggest learning prospectively to prepare for future updates, and propose ForwArd Compatible Training (FACT) for FSCIL. Forward compatibility requires future new classes to be easily incorporated into the current model based on the current stage data, and we seek to realize it by reserving embedding space for future new classes. In detail, we assign virtual prototypes to squeeze the embedding of known classes and reserve for new ones. Besides, we forecast possible new classes and prepare for the updating process. The virtual prototypes allow the model to accept possible updates in the future, which act as proxies scattered among embedding space to build a stronger classifier during inference. FACT efficiently incorporates new classes with forward compatibility and meanwhile resists forgetting of old ones. Extensive experiments validate FACT's state-of-the-art performance. Code is available at: https://github.com/zhoudw-zdw/CVPR22-Fact

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Self-distilled Knowledge Delegator for Exemplar-free Class Incremental Learning

Exemplar-free incremental learning is extremely challenging due to inaccessibility of data from old tasks. In this paper, we attempt to exploit the knowledge encoded in a previously trained classification model to handle the catastrophic forgetting problem in continual learning. Specifically, we introduce a so-called knowledge delegator, which is capable of transferring knowledge from the trained model to a randomly re-initialized new model by generating informative samples. Given the previous model only, the delegator is effectively learned using a self-distillation mechanism in a data-free manner. The knowledge extracted by the delegator is then utilized to maintain the performance of the model on old tasks in incremental learning. This simple incremental learning framework surpasses existing exemplar-free methods by a large margin on four widely used class incremental benchmarks, namely CIFAR-100, ImageNet-Subset, Caltech-101 and Flowers-102. Notably, we achieve comparable performance to some exemplar-based methods without accessing any exemplars.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Making artificial $\textit{p}_{x,y}$-orbital honeycomb electron lattice on metal surface

We theoretically demonstrate that the desired $p_{x,y}$-orbital honeycomb electron lattice can be readily realized by arranging CO molecules into a hexagonal lattice on Cu(111) surface with scanning tunneling microscopy (STM). The electronic structure of the Cu surface states in the presence of CO molecules is calculated with various methods, \textit{i.e.}~DFT simulations, muffin-tin potential model and tight-binding model. Our calculations indicate that, by measuring the LDOS pattern using STM, the $p$-orbital surface bands can be immediately identified in experiment. We also give an analytic interpretation of the $p$-orbital LDOS pattern with $k \cdot p$ method. Meanwhile, different from the case of graphene, the $p$-orbital honeycomb lattice has two kinds of edge states, which can also be directly observed in STM experiment. Our work points out a feasible way to construct a $p_{x,y}$-orbital honeycomb electron lattice in a real system, which may have exotic properties, such as Wigner crystal, ferromagnetism, $f$-wave superconductivity, quantum anomalous Hall (QAH) effect. Furthermore, we also propose a simple way to calculate and identify the modified Cu surface bands in the Cu/CO systems with the DFT simulations. Considering the recent works about $p$-orbital square lattice in similar systems [M. R. Slot, \textit{et al.} Nat. Phys. \textbf{13}, 672 (2017); Liang Ma, \textit{et al.} Phys. Rev. B \textbf{99}, 205403 (2019)], our work once again illustrates that the artificial electron lattice on metal surface is an ideal platform to study the orbital physics in a controllable way.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Experimental Syntheses of Sodalite-like Clathrate EuH$_6$ and EuH$_9$ at Extreme Pressures

The recent discovery of a class of sodalite-like clathrate superhydrides (e.g., YH6, YH9, ThH9, ThH10, and LaH10) at extreme pressures, which exhibit commonly a high-temperature superconductivity with the highest Tc approaching 260 K for LaH10, opened up a new era in search of high-temperature superconductors in metal superhydrides. There is a high interest towards the finding of alternative clathrate superhydrides that might witness the long-dreamed room-temperature superconductivity. Here, we target on the experimental synthesis of strongly-correlated europium (Eu) superhydrides where theory can fail for the prediction of superconductivity. We pressurized and laser-heated the mixture of metal Eu and ammonia borane (NH3BH3) in a diamond anvil cell and successfully synthesized the sodalite-like clathrate EuH6 and EuH9 at conditions of 152 GPa and 1,700 K, and 170 GPa and 2,800 K, respectively. Two non-clathrate structured phases of EuH5 and EuH6 were also synthesized that are not reported in lanthanide superhydrides. Calculated large H-derived electronic density of states at the Fermi level in clathrate EuH6 implies the potential of high temperature superconductivity. Our work created a model superhydride platform for subsequent investigation on how strongly-correlated effect in electronic structure can affect the superconductivity of superhydrides, a phenomenon that is not known thus far.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

IROS 2019 Lifelong Robotic Vision Challenge -- Lifelong Object Recognition Report

This report summarizes IROS 2019-Lifelong Robotic Vision Competition (Lifelong Object Recognition Challenge) with methods and results from the top $8$ finalists (out of over~$150$ teams). The competition dataset (L)ifel(O)ng (R)obotic V(IS)ion (OpenLORIS) - Object Recognition (OpenLORIS-object) is designed for driving lifelong/continual learning research and application in robotic vision domain, with everyday objects in home, office, campus, and mall scenarios. The dataset explicitly quantifies the variants of illumination, object occlusion, object size, camera-object distance/angles, and clutter information. Rules are designed to quantify the learning capability of the robotic vision system when faced with the objects appearing in the dynamic environments in the contest. Individual reports, dataset information, rules, and released source code can be found at the project homepage: "https://lifelong-robotic-vision.github.io/competition/".

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Neural Network Tomography

Network tomography, a classic research problem in the realm of network monitoring, refers to the methodology of inferring unmeasured network attributes using selected end-to-end path measurements. In the research community, network tomography is generally investigated under the assumptions of known network topology, correlated path measurements, bounded number of faulty nodes/links, or even special network protocol support. The applicability of network tomography is considerably constrained by these strong assumptions, which therefore frequently position it in the theoretical world. In this regard, we revisit network tomography from the practical perspective by establishing a generic framework that does not rely on any of these assumptions or the types of performance metrics. Given only the end-to-end path performance metrics of sampled node pairs, the proposed framework, NeuTomography, utilizes deep neural network and data augmentation to predict the unmeasured performance metrics via learning non-linear relationships between node pairs and underlying unknown topological/routing properties. In addition, NeuTomography can be employed to reconstruct the original network topology, which is critical to most network planning tasks. Extensive experiments using real network data show that comparing to baseline solutions, NeuTomography can predict network characteristics and reconstruct network topologies with significantly higher accuracy and robustness using only limited measurement data.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

State Action Separable Reinforcement Learning

Reinforcement Learning (RL) based methods have seen their paramount successes in solving serial decision-making and control problems in recent years. For conventional RL formulations, Markov Decision Process (MDP) and state-action-value function are the basis for the problem modeling and policy evaluation. However, several challenging issues still remain. Among most cited issues, the enormity of state/action space is an important factor that causes inefficiency in accurately approximating the state-action-value function. We observe that although actions directly define the agents' behaviors, for many problems the next state after a state transition matters more than the action taken, in determining the return of such a state transition. In this regard, we propose a new learning paradigm, State Action Separable Reinforcement Learning (sasRL), wherein the action space is decoupled from the value function learning process for higher efficiency. Then, a light-weight transition model is learned to assist the agent to determine the action that triggers the associated state transition. In addition, our convergence analysis reveals that under certain conditions, the convergence time of sasRL is $O(T^{1/k})$, where $T$ is the convergence time for updating the value function in the MDP-based formulation and $k$ is a weighting factor. Experiments on several gaming scenarios show that sasRL outperforms state-of-the-art MDP-based RL algorithms by up to $75\%$.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Synetgy: Algorithm-hardware Co-design for ConvNet Accelerators on Embedded FPGAs

Using FPGAs to accelerate ConvNets has attracted significant attention in recent years. However, FPGA accelerator design has not leveraged the latest progress of ConvNets. As a result, the key application characteristics such as frames-per-second (FPS) are ignored in favor of simply counting GOPs, and results on accuracy, which is critical to application success, are often not even reported. In this work, we adopt an algorithm-hardware co-design approach to develop a ConvNet accelerator called Synetgy and a novel ConvNet model called DiracDeltaNet$^{\dagger}$. Both the accelerator and ConvNet are tailored to FPGA requirements. DiracDeltaNet, as the name suggests, is a ConvNet with only $1\times 1$ convolutions while spatial convolutions are replaced by more efficient shift operations. DiracDeltaNet achieves competitive accuracy on ImageNet (88.7\% top-5), but with 42$\times$ fewer parameters and 48$\times$ fewer OPs than VGG16. We further quantize DiracDeltaNet's weights to 4-bit and activations to 4-bits, with less than 1\% accuracy loss. These quantizations exploit well the nature of FPGA hardware. In short, DiracDeltaNet's small model size, low computational OP count, low precision and simplified operators allow us to co-design a highly customized computing unit for an FPGA. We implement the computing units for DiracDeltaNet on an Ultra96 SoC system through high-level synthesis. Our accelerator's final top-5 accuracy of 88.1\% on ImageNet, is higher than all the previously reported embedded FPGA accelerators. In addition, the accelerator reaches an inference speed of 66.3 FPS on the ImageNet classification task, surpassing prior works with similar accuracy by at least 11.6$\times$.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Vacua and Exact Solutions in Lower-$D$ Limits of EGB

We consider the action principles that are the lower dimensional limits of the Einstein-Gauss-Bonnet gravity {\it via} the Kaluza-Klein route. We study the vacua and obtain some exact solutions. We find that the reality condition of the theories may select one vacuum over the other from the two vacua that typically arise in Einstein-Gauss-Bonnet gravity. We obtain exact black hole and cosmological solutions carrying scalar hair, including scalar hairy BTZ black holes with both mass and angular momentum turned on. We also discuss the holographic central charges in the asymptotic AdS backgrounds.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Bounds on Photon Spheres and Shadows of Charged Black Holes in Einstein-Gauss-Bonnet-Maxwell Gravity

We consider spherically symmetric and static charged black holes in Einstein-Gauss-Bonnet-Maxwell gravities in general $D\ge 5$ dimensions and study their photon spheres and black hole shadows. We show that they all satisfy the sequence of inequalities recently proposed relating a black hole's horizon, photon sphere, shadow and its mass.

0 citations0 reviewsNo code linkedOpen access
preprint2019arXiv

Holographic Complexity Bounds

We study the action growth rate in the Wheeler-DeWitt (WDW) patch for a variety of $D\ge 4$ black holes in Einstein gravity that are asymptotic to the anti-de Sitter spacetime, with spherical, toric and hyperbolic horizons, corresponding to the topological parameter $k=1,0,-1$ respectively. We find a lower bound inequality $\frac{1}{T} \frac{\partial \dot I_{\rm WDW}}{\partial S}|_{Q,P_{\rm th}}> C$ for $k=0,1$, where $C$ is some order-one numerical constant. The lowest number in our examples is $C=(D-3)/(D-2)$. We also find that the quantity $(\dot I_{\rm WDW}-2P_{\rm th}\, ΔV_{\rm th})$ is greater than, equal to, or less than zero, for $k=1,0,-1$ respectively. For black holes with two horizons, $ΔV_{\rm th}=V_{\rm th}^+-V_{\rm th}^-$, i.e. the difference between the thermodynamical volumes of the outer and inner horizons. For black holes with only one horizon, we introduce a new concept of the volume $V_{\rm th}^0$ of the black hole singularity, and define $ΔV_{\rm th}=V_{\rm th}^+-V_{\rm th}^0$. The volume $V_{\rm th}^0$ vanishes for the Schwarzschild black hole, but in general it can be positive, negative or even divergent. For black holes with single horizon, we find a relation between $\dot I_{\rm WDW}$ and $V_{\rm th}^0$, which implies that the holographic complexity preserves the Lloyd's bound for positive or vanishing $V_{\rm th}^0$, but the bound is violated when $V_{\rm th}^0$ becomes negative. We also find explicit black hole examples where $V_{\rm th}^0$ and hence $\dot I_{\rm WDW}$ are divergent.

0 citations0 reviewsNo code linkedOpen access