arXiv subjects

Tao Lin

Publications and source records attributed to Tao Lin.

At least 145 records · Page 8Linked to original sources

arXiv2024

Open-Source AI-based SE Tools: Opportunities and Challenges of Collaborative Software Learning

Large Language Models (LLMs) have become instrumental in advancing software engineering (SE) tasks, showcasing their efficacy in code understanding and beyond. Like traditional SE tools, open-source collaboration is key in realising the excellent products. However, with AI models, the essential need is in data. The collaboration of these AI-based SE models hinges on maximising the sources of high-quality data. However, data especially of high quality, often holds commercial or sensitive value, making it less accessible for open-source AI-based SE projects. This reality presents a significant barrier to the development and enhancement of AI-based SE tools within the software engineering community. Therefore, researchers need to find solutions for enabling open-source AI-based SE models to tap into resources by different organisations. Addressing this challenge, our position paper investigates one solution to facilitate access to diverse organizational resources for open-source AI models, ensuring privacy and commercial sensitivities are respected. We introduce a governance framework centered on federated learning (FL), designed to foster the joint development and maintenance of open-source AI code models while safeguarding data privacy and security. Additionally, we present guidelines for developers on AI-based SE tool collaboration, covering data requirements, model architecture, updating strategies, and version control. Given the significant influence of data characteristics on FL, our research examines the effect of code data heterogeneity on FL performance.

cs.SE↗

arXiv2024

Towards Robust Multi-Modal Reasoning via Model Selection

The reasoning capabilities of LLM (Large Language Model) are widely acknowledged in recent research, inspiring studies on tool learning and autonomous agents. LLM serves as the "brain" of the agent, orchestrating multiple tools for collaborative multi-step task solving. Unlike methods invoking tools like calculators or weather APIs for straightforward tasks, multi-modal agents excel by integrating diverse AI models for complex challenges. However, current multi-modal agents neglect the significance of model selection: they primarily focus on the planning and execution phases, and will only invoke predefined task-specific models for each subtask, making the execution fragile. Meanwhile, other traditional model selection methods are either incompatible with or suboptimal for the multi-modal agent scenarios, due to ignorance of dependencies among subtasks arising by multi-step reasoning. To this end, we identify the key challenges therein and propose the $\textit{M}^3$ framework as a plug-in with negligible runtime overhead at test-time. This framework improves model selection and bolsters the robustness of multi-modal agents in multi-step reasoning. In the absence of suitable benchmarks, we create MS-GQA, a new dataset specifically designed to investigate the model selection challenge in multi-modal agents. Our experiments reveal that our framework enables dynamic model selection, considering both user inputs and subtask dependencies, thereby robustifying the overall reasoning process. Our code and benchmark: https://github.com/LINs-lab/M3.

cs.LG↗

arXiv2024

PathMMU: A Massive Multimodal Expert-Level Benchmark for Understanding and Reasoning in Pathology

The emergence of large multimodal models has unlocked remarkable potential in AI, particularly in pathology. However, the lack of specialized, high-quality benchmark impeded their development and precise evaluation. To address this, we introduce PathMMU, the largest and highest-quality expert-validated pathology benchmark for Large Multimodal Models (LMMs). It comprises 33,428 multimodal multi-choice questions and 24,067 images from various sources, each accompanied by an explanation for the correct answer. The construction of PathMMU harnesses GPT-4V's advanced capabilities, utilizing over 30,000 image-caption pairs to enrich captions and generate corresponding Q&As in a cascading process. Significantly, to maximize PathMMU's authority, we invite seven pathologists to scrutinize each question under strict standards in PathMMU's validation and test sets, while simultaneously setting an expert-level performance benchmark for PathMMU. We conduct extensive evaluations, including zero-shot assessments of 14 open-sourced and 4 closed-sourced LMMs and their robustness to image corruption. We also fine-tune representative LMMs to assess their adaptability to PathMMU. The empirical findings indicate that advanced LMMs struggle with the challenging PathMMU benchmark, with the top-performing LMM, GPT-4V, achieving only a 49.8% zero-shot performance, significantly lower than the 71.8% demonstrated by human pathologists. After fine-tuning, significantly smaller open-sourced LMMs can outperform GPT-4V but still fall short of the expertise shown by pathologists. We hope that the PathMMU will offer valuable insights and foster the development of more specialized, next-generation LMMs for pathology.

cs.CV↗

arXiv2024

On the Diversity and Realism of Distilled Dataset: An Efficient Dataset Distillation Paradigm

Contemporary machine learning requires training large neural networks on massive datasets and thus faces the challenges of high computational demands. Dataset distillation, as a recent emerging strategy, aims to compress real-world datasets for efficient training. However, this line of research currently struggle with large-scale and high-resolution datasets, hindering its practicality and feasibility. To this end, we re-examine the existing dataset distillation methods and identify three properties required for large-scale real-world applications, namely, realism, diversity, and efficiency. As a remedy, we propose RDED, a novel computationally-efficient yet effective data distillation paradigm, to enable both diversity and realism of the distilled data. Extensive empirical results over various neural architectures and datasets demonstrate the advancement of RDED: we can distill the full ImageNet-1K to a small dataset comprising 10 images per class within 7 minutes, achieving a notable 42% top-1 accuracy with ResNet-18 on a single RTX-4090 GPU (while the SOTA only achieves 21% but requires 6 hours).

cs.CV↗

arXiv2024

Modular Multi-Level Replanning TAMP Framework for Dynamic Environment

Task and Motion Planning (TAMP) algorithms can generate plans that combine logic and motion aspects for robots. However, these plans are sensitive to interference and control errors. To make TAMP more applicable in real-world, we propose the modular multi-level replanning TAMP framework(MMRF), blending the probabilistic completeness of sampling-based TAMP algorithm with the robustness of reactive replanning. MMRF generates an nominal plan from the initial state, then dynamically reconstructs this nominal plan in real-time, reorders robot manipulations. Following the logic-level adjustment, GMRF will try to replan a new motion path to ensure the updated plan is feasible at the motion level. Finally, we conducted real-world experiments involving stack and rearrange task domains. The result demonstrate MMRF's ability to swiftly complete tasks in scenarios with varying degrees of interference.

cs.RO↗

arXiv2024

Persuading a Behavioral Agent: Approximately Best Responding and Learning

The classic Bayesian persuasion model assumes a Bayesian and best-responding receiver. We study a relaxation of the Bayesian persuasion model where the receiver can approximately best respond to the sender's signaling scheme. We show that, under natural assumptions, (1) the sender can find a signaling scheme that guarantees itself an expected utility almost as good as its optimal utility in the classic model, no matter what approximately best-responding strategy the receiver uses; (2) on the other hand, there is no signaling scheme that gives the sender much more utility than its optimal utility in the classic model, even if the receiver uses the approximately best-responding strategy that is best for the sender. Together, (1) and (2) imply that the approximately best-responding behavior of the receiver does not affect the sender's maximal achievable utility a lot in the Bayesian persuasion problem. The proofs of both results rely on the idea of robustification of a Bayesian persuasion scheme: given a pair of the sender's signaling scheme and the receiver's strategy, we can construct another signaling scheme such that the receiver prefers to use that strategy in the new scheme more than in the original scheme, and the two schemes give the sender similar utilities. As an application of our main result (1), we show that, in a repeated Bayesian persuasion model where the receiver learns to respond to the sender by some algorithms, the sender can do almost as well as in the classic model. Interestingly, unlike (2), with a learning receiver the sender can sometimes do much better than in the classic model.

cs.GT↗

arXiv2024

Reconstructing parton distribution function based on maximum entropy method

A new method based on the maximum entropy principle for reconstructing the parton distribution function (PDF) from moments is proposed. Unlike traditional methods, the new method no longer needs to introduce any artificial assumptions. For the case of moments with errors, we introduce Gaussian functions to soften the constraints of moments. Through a series of tests, the effectiveness and reconstruction efficiency of this new method are evaluated comprehensively. And these tests indicate that this method is reasonable and can achieve high-quality reconstruction with at least the first six moments as input. Finally, we select a set of lattice QCD results regarding moments as input and provide reasonable reconstruction results for the pion.

hep-ph↗

arXiv2024

A Data-driven dE/dx Simulation with Normalizing Flow

In high-energy physics, precise measurements rely on highly reliable detector simulations. Traditionally, these simulations involve incorporating experiment data to model detector responses and fine-tuning them. However, due to the complexity of the experiment data, tuning the simulation can be challenging. One crucial aspect for charged particle identification is the measurement of energy deposition per unit length (referred to as dE/dx). This paper proposes a data-driven dE/dx simulation method using the Normalizing Flow technique, which can learn the dE/dx distribution directly from experiment data. By employing this method, not only can the need for manual tuning of the dE/dx simulation be eliminated, but also high-precision simulation can be achieved.

hep-ex↗

arXiv2023

A High Order Geometry Conforming Immersed Finite Element for Elliptic Interface Problems

We present a high order immersed finite element (IFE) method for solving the elliptic interface problem with interface-independent meshes. The IFE functions developed here satisfy the interface conditions exactly and they have optimal approximation capabilities. The construction of this novel IFE space relies on a nonlinear transformation based on the Frenet-Serret frame of the interface to locally map it into a line segment, and this feature makes the process of constructing the IFE functions cost-effective and robust for any degree. This new class of immersed finite element functions is locally conforming with the usual weak form of the interface problem so that they can be employed in the standard interior penalty discontinuous Galerkin scheme without additional penalties on the interface. Numerical examples are provided to showcase the convergence properties of the method under $h$ and $p$ refinements.

math.NA↗

arXiv2023

The Key4hep software stack: Beyond Future Higgs factories

The Key4hep project aims to provide a turnkey software solution for the full experiment lifecycle, based on established community tools. Several future collider communities (CEPC, CLIC, EIC, FCC, and ILC) have joined to develop and adapt their workflows to use the common data model EDM4hep and common framework. Besides sharing of existing experiment workflows, one focus of the Key4hep project is the development and integration of new experiment independent software libraries. Ongoing collaborations with projects such as ACTS, CLUE, PandoraPFA and the OpenDataDector show the potential of Key4hep as an experiment-independent testbed and development platform. In this talk, we present the challenges of an experiment-independent framework along with the lessons learned from discussions of interested communities (such as LUXE) and recent adopters of Key4hep in order to discuss how Key4hep could be of interest to the wider HEP community while staying true to its goal of supporting future collider designs studies.

hep-ex↗

arXiv2023

Key4hep: Progress Report on Integrations

Detector studies for future experiments rely on advanced software tools to estimate performance and optimize their design and technology choices. The Key4hep project provides a flexible turnkey solution for the full experiment life-cycle based on established community tools such as ROOT, Geant4, DD4hep, Gaudi, podio and spack. Members of the CEPC, CLIC, EIC, FCC, and ILC communities have joined to develop this framework and have merged, or are in the progress of merging, their respective software environments into the Key4hep stack. These proceedings will give an overview over the recent progress in the Key4hep project: covering the developments towards adaptation of state-of-the-art tools for simulation (DD4hep, Gaussino), track and calorimeter reconstruction (ACTS, CLUE), particle flow (PandoraPFA), analysis via RDataFrame, and visualization with Phoenix, as well as tools for testing and validation.

hep-ex↗

arXiv2023

Learning Thresholds with Latent Values and Censored Feedback

In this paper, we investigate a problem of actively learning threshold in latent space, where the unknown reward $g(γ, v)$ depends on the proposed threshold $γ$ and latent value $v$ and it can be $only$ achieved if the threshold is lower than or equal to the unknown latent value. This problem has broad applications in practical scenarios, e.g., reserve price optimization in online auctions, online task assignments in crowdsourcing, setting recruiting bars in hiring, etc. We first characterize the query complexity of learning a threshold with the expected reward at most $ε$ smaller than the optimum and prove that the number of queries needed can be infinitely large even when $g(γ, v)$ is monotone with respect to both $γ$ and $v$. On the positive side, we provide a tight query complexity $\tildeΘ(1/ε^3)$ when $g$ is monotone and the CDF of value distribution is Lipschitz. Moreover, we show a tight $\tildeΘ(1/ε^3)$ query complexity can be achieved as long as $g$ satisfies one-sided Lipschitzness, which provides a complete characterization for this problem. Finally, we extend this model to an online learning setting and demonstrate a tight $Θ(T^{2/3})$ regret bound using continuous-arm bandit techniques and the aforementioned query complexity results.

cs.LG↗

arXiv2023

Networked Time Series Prediction with Incomplete Data via Generative Adversarial Network

A networked time series (NETS) is a family of time series on a given graph, one for each node. It has a wide range of applications from intelligent transportation, environment monitoring to smart grid management. An important task in such applications is to predict the future values of a NETS based on its historical values and the underlying graph. Most existing methods require complete data for training. However, in real-world scenarios, it is not uncommon to have missing data due to sensor malfunction, incomplete sensing coverage, etc. In this paper, we study the problem of NETS prediction with incomplete data. We propose NETS-ImpGAN, a novel deep learning framework that can be trained on incomplete data with missing values in both history and future. Furthermore, we propose Graph Temporal Attention Networks, which incorporate the attention mechanism to capture both inter-time series and temporal correlations. We conduct extensive experiments on four real-world datasets under different missing patterns and missing rates. The experimental results show that NETS-ImpGAN outperforms existing methods, reducing the MAE by up to 25%.

cs.LG↗

arXiv2023

DELTA: Diverse Client Sampling for Fasting Federated Learning

Partial client participation has been widely adopted in Federated Learning (FL) to reduce the communication burden efficiently. However, an inadequate client sampling scheme can lead to the selection of unrepresentative subsets, resulting in significant variance in model updates and slowed convergence. Existing sampling methods are either biased or can be further optimized for faster convergence.In this paper, we present DELTA, an unbiased sampling scheme designed to alleviate these issues. DELTA characterizes the effects of client diversity and local variance, and samples representative clients with valuable information for global model updates. In addition, DELTA is a proven optimal unbiased sampling scheme that minimizes variance caused by partial client participation and outperforms other unbiased sampling schemes in terms of convergence. Furthermore, to address full-client gradient dependence,we provide a practical version of DELTA depending on the available clients' information, and also analyze its convergence. Our results are validated through experiments on both synthetic and real-world datasets.

cs.LG↗

arXiv2023

Revisiting Implicit Models: Sparsity Trade-offs Capability in Weight-tied Model for Vision Tasks

Implicit models such as Deep Equilibrium Models (DEQs) have garnered significant attention in the community for their ability to train infinite layer models with elegant solution-finding procedures and constant memory footprint. However, despite several attempts, these methods are heavily constrained by model inefficiency and optimization instability. Furthermore, fair benchmarking across relevant methods for vision tasks is missing. In this work, we revisit the line of implicit models and trace them back to the original weight-tied models. Surprisingly, we observe that weight-tied models are more effective, stable, as well as efficient on vision tasks, compared to the DEQ variants. Through the lens of these simple-yet-clean weight-tied models, we further study the fundamental limits in the model capacity of such models and propose the use of distinct sparse masks to improve the model capacity. Finally, for practitioners, we offer design guidelines regarding the depth, width, and sparsity selection for weight-tied models, and demonstrate the generalizability of our insights to other learning paradigms.

cs.LG↗

arXiv2023

Sample Complexity of Forecast Aggregation

We consider a Bayesian forecast aggregation model where $n$ experts, after observing private signals about an unknown binary event, report their posterior beliefs about the event to a principal, who then aggregates the reports into a single prediction for the event. The signals of the experts and the outcome of the event follow a joint distribution that is unknown to the principal, but the principal has access to i.i.d. "samples" from the distribution, where each sample is a tuple of the experts' reports (not signals) and the realization of the event. Using these samples, the principal aims to find an $\varepsilon$-approximately optimal aggregator, where optimality is measured in terms of the expected squared distance between the aggregated prediction and the realization of the event. We show that the sample complexity of this problem is at least $\tilde Ω(m^{n-2} / \varepsilon)$ for arbitrary discrete distributions, where $m$ is the size of each expert's signal space. This sample complexity grows exponentially in the number of experts $n$. But, if the experts' signals are independent conditioned on the realization of the event, then the sample complexity is significantly reduced, to $\tilde O(1 / \varepsilon^2)$, which does not depend on $n$. Our results can be generalized to non-binary events. The proof of our results uses a reduction from the distribution learning problem and reveals the fact that forecast aggregation is almost as difficult as distribution learning.

cs.LG↗

arXiv2023

Modern Software Development for JUNO offline software

The Jiangmen Underground Neutrino Observatory (JUNO), under construction in South China, primarily aims to determine the neutrino mass hierarchy and to precise measure the neutrino oscillation parameters. The data-taking is expected to start in 2024 and the detector plans to run for more than 20 years. The development of the JUNO offline software (JUNOSW) started in 2012, and it is quite challenging to maintain the JUNOSW for such a long time. In the last ten years, tools such as Subversion, Trac, and CMT had been adopted for software development. However, new stringent requirements came out, such as how to reduce the building time for the whole project, how to deploy offline algorithms to an online environment, and how to improve the code quality with code review and continuous integration. To meet the further requirements of software development, modern development tools are evaluated for JUNOSW, such as Git, GitLab, CMake, Docker, and Kubernetes. This contribution will present the software development system based on these modern tools for JUNOSW and the functionalities achieved: CMake macros are developed to simplify the build instructions for users; CMake generator expressions are used to control the build flags for the online and offline environments; a tool named git-junoenv is developed to help users partially checkout and build the software; a script is used to build and deploy the software on the CVMFS server; a Docker image with CVMFS client installed is created for continuous integration; a GitLab agent is set up to manage GitLab runners in Kubernetes with all the configurations in a GitLab repository.

physics.ins-det↗

arXiv2023

No Fear of Classifier Biases: Neural Collapse Inspired Federated Learning with Synthetic and Fixed Classifier

Data heterogeneity is an inherent challenge that hinders the performance of federated learning (FL). Recent studies have identified the biased classifiers of local models as the key bottleneck. Previous attempts have used classifier calibration after FL training, but this approach falls short in improving the poor feature representations caused by training-time classifier biases. Resolving the classifier bias dilemma in FL requires a full understanding of the mechanisms behind the classifier. Recent advances in neural collapse have shown that the classifiers and feature prototypes under perfect training scenarios collapse into an optimal structure called simplex equiangular tight frame (ETF). Building on this neural collapse insight, we propose a solution to the FL's classifier bias problem by utilizing a synthetic and fixed ETF classifier during training. The optimal classifier structure enables all clients to learn unified and optimal feature representations even under extremely heterogeneous data. We devise several effective modules to better adapt the ETF structure in FL, achieving both high generalization and personalization. Extensive experiments demonstrate that our method achieves state-of-the-art performances on CIFAR-10, CIFAR-100, and Tiny-ImageNet.

cs.LG↗