arXiv subjects

Lu Yin

Publications and source records attributed to Lu Yin.

At least 91 records · Page 5Linked to original sources

arXiv2024

Multimodal Contrastive Learning of Urban Space Representations from POI Data

Existing methods for learning urban space representations from Point-of-Interest (POI) data face several limitations, including issues with geographical delineation, inadequate spatial information modelling, underutilisation of POI semantic attributes, and computational inefficiencies. To address these issues, we propose CaLLiPer (Contrastive Language-Location Pre-training), a novel representation learning model that directly embeds continuous urban spaces into vector representations that can capture the spatial and semantic distribution of urban environment. This model leverages a multimodal contrastive learning objective, aligning location embeddings with textual POI descriptions, thereby bypassing the need for complex training corpus construction and negative sampling. We validate CaLLiPer's effectiveness by applying it to learning urban space representations in London, UK, where it demonstrates 5-15% improvement in predictive performance for land use classification and socioeconomic mapping tasks compared to state-of-the-art methods. Visualisations of the learned representations further illustrate our model's advantages in capturing spatial variations in urban semantics with high accuracy and fine resolution. Additionally, CaLLiPer achieves reduced training time, showcasing its efficiency and scalability. This work provides a promising pathway for scalable, semantically rich urban space representation learning that can support the development of geospatial foundation models. The implementation code is available at https://github.com/xlwang233/CaLLiPer.

cs.AI↗

arXiv2024

Full-Rank No More: Low-Rank Weight Training for Modern Speech Recognition Models

This paper investigates the under-explored area of low-rank weight training for large-scale Conformer-based speech recognition models from scratch. Our study demonstrates the viability of this training paradigm for such models, yielding several notable findings. Firstly, we discover that applying a low-rank structure exclusively to the attention modules can unexpectedly enhance performance, even with a significant rank reduction of 12%. In contrast, feed-forward layers present greater challenges, as they begin to exhibit performance degradation with a moderate 50% rank reduction. Furthermore, we find that both initialization and layer-wise rank assignment play critical roles in successful low-rank training. Specifically, employing SVD initialization and linear layer-wise rank mapping significantly boosts the efficacy of low-rank weight training. Building on these insights, we introduce the Low-Rank Speech Model from Scratch (LR-SMS), an approach that achieves performance parity with full-rank training while delivering substantial reductions in parameters count (by at least 2x), and training time speedups (by 1.3x for ASR and 1.15x for AVSR).

cs.SD↗

arXiv2024

Is C4 Dataset Optimal for Pruning? An Investigation of Calibration Data for LLM Pruning

Network pruning has emerged as a potential solution to make LLMs cheaper to deploy. However, existing LLM pruning approaches universally rely on the C4 dataset as the calibration data for calculating pruning scores, leaving its optimality unexplored. In this study, we evaluate the choice of calibration data on LLM pruning, across a wide range of datasets that are most commonly used in LLM training and evaluation, including four pertaining datasets as well as three categories of downstream tasks encompassing nine datasets. Each downstream dataset is prompted with In-Context Learning (ICL) and Chain-of-Thought (CoT), respectively. Besides the already intriguing observation that the choice of calibration data significantly impacts the performance of pruned LLMs, our results also uncover several subtle and often unexpected findings, summarized as follows: (1) C4 is not the optimal choice for LLM pruning, even among commonly used pre-training datasets; (2) arithmetic datasets, when used as calibration data, performs on par or even better than pre-training datasets; (3) pruning with downstream datasets does not necessarily help the corresponding downstream task, compared to pre-training data; (4) ICL is widely beneficial to all data categories, whereas CoT is only useful on certain tasks. Our findings shed light on the importance of carefully selecting calibration data for LLM pruning and pave the way for more efficient deployment of these powerful models in real-world applications. We release our code at: https://github.com/abx393/llm-pruning-calibration-data.

cs.CL↗

arXiv2024

Gauss-Bonnet Cosmology: large-temperature behaviour and bounds from Gravitational Waves

We provide a transparent discussion of the high temperature asymptotic behaviour of Cosmology in a dilaton-Einstein-Gauss-Bonnet (dEGB) scenario of modified gravity with vanishing scalar potential. In particular, we show that it has a clear interpretation in terms of only three attractors (stable critical points) of a set of autonomous differential equations: $w=-\frac{1}{3}$, $w=1$ and $1<w<\frac{7}{3}$, where $w\equiv p/ρ$ is the equation of state, defined as the ratio of the total pressure and the total energy density. All the possible different high-temperature evolution histories of the model are exhausted by only eight paths in the flow of the set of the autonomous differential equations. Our discussion clearly explains why five out of them are characterized by a swift transition of the system toward the attractor, while the remaining three show a more convoluted evolution, where the system follows a meta-stable equation of state at intermediate temperatures before eventually jumping to the real attractor at higher temperatures. Compared to standard Cosmology, the regions of the dEGB parameter space with $w=-\frac{1}{3}$ show a strong enhancement of the expected Gravitational Wave stochastic background produced by the primordial plasma of relativistic particles of the Standard Model. This is due to the very peculiar fact that dEGB allows to have an epoch when the energy density $ρ_{\rm rad}$ of the relativistic plasma dominates the energy of the Universe while at the same time the rate of dilution with $T$ of the total energy density is slower than what usually expected during radiation dominance. This allows to use the bound from BBN to put in dEGB a constraint $T_{\rm RH}\lesssim 10^8 - 10^9$ GeV on the reheating temperature of the Universe $T_{\rm RH}$. Such BBN bound is complementary to late-time constraints from compact binary mergers.

hep-ph↗

arXiv2024

Robust Active Learning (RoAL): Countering Dynamic Adversaries in Active Learning with Elastic Weight Consolidation

Despite significant advancements in active learning and adversarial attacks, the intersection of these two fields remains underexplored, particularly in developing robust active learning frameworks against dynamic adversarial threats. The challenge of developing robust active learning frameworks under dynamic adversarial attacks is critical, as these attacks can lead to catastrophic forgetting within the active learning cycle. This paper introduces Robust Active Learning (RoAL), a novel approach designed to address this issue by integrating Elastic Weight Consolidation (EWC) into the active learning process. Our contributions are threefold: First, we propose a new dynamic adversarial attack that poses significant threats to active learning frameworks. Second, we introduce a novel method that combines EWC with active learning to mitigate catastrophic forgetting caused by dynamic adversarial attacks. Finally, we conduct extensive experimental evaluations to demonstrate the efficacy of our approach. The results show that RoAL not only effectively counters dynamic adversarial threats but also significantly reduces the impact of catastrophic forgetting, thereby enhancing the robustness and performance of active learning systems in adversarial environments.

cs.LG↗

arXiv2024

Q-GaLore: Quantized GaLore with INT4 Projection and Layer-Adaptive Low-Rank Gradients

Training Large Language Models (LLMs) is memory-intensive due to the large number of parameters and associated optimization states. GaLore, a recent method, reduces memory usage by projecting weight gradients into a low-rank subspace without compromising performance. However, GaLore relies on time-consuming Singular Value Decomposition (SVD) operations to identify the subspace, and the frequent subspace updates lead to significant training time overhead. Moreover, GaLore offers minimal improvements in accuracy and efficiency compared to LoRA in more accessible fine-tuning scenarios. To address these limitations, we introduce Q-Galore, a novel approach that substantially reduces memory usage by combining quantization and low-rank projection, surpassing the benefits of GaLore. Our method is based on two key observations: (i) the gradient subspace exhibits diverse properties, with some layers converging early in training while others are subject to frequent changes; (ii) the projection matrices are highly resilient to low-bit quantization. Leveraging these insights, Q-GaLore adaptively updates the gradient subspace based on its convergence statistics, achieving comparable performance while significantly reducing the number of SVD operations. We maintain the projection matrices in INT4 format and weights in INT8 format, incorporating stochastic rounding to capture accumulated gradient information. This approach enables a high-precision training trajectory using only low-precision weights. We demonstrate that Q-GaLore achieves highly competitive performance with exceptional memory efficiency. At pre-training, Q-GaLore facilitates training a LLaMA-7B model from scratch on a single NVIDIA RTX 4060 Ti with only 16 GB memory. At fine-tuning, it reduces memory consumption by up to 50% compared to LoRA and GaLore, while consistently outperforming QLoRA at the same memory cost.

cs.LG↗

arXiv2024

Dynamic Data Pruning for Automatic Speech Recognition

The recent success of Automatic Speech Recognition (ASR) is largely attributed to the ever-growing amount of training data. However, this trend has made model training prohibitively costly and imposed computational demands. While data pruning has been proposed to mitigate this issue by identifying a small subset of relevant data, its application in ASR has been barely explored, and existing works often entail significant overhead to achieve meaningful results. To fill this gap, this paper presents the first investigation of dynamic data pruning for ASR, finding that we can reach the full-data performance by dynamically selecting 70% of data. Furthermore, we introduce Dynamic Data Pruning for ASR (DDP-ASR), which offers several fine-grained pruning granularities specifically tailored for speech-related datasets, going beyond the conventional pruning of entire time sequences. Our intensive experiments show that DDP-ASR can save up to 1.6x training time with negligible performance loss.

cs.CL↗

arXiv2024

MSRS: Training Multimodal Speech Recognition Models from Scratch with Sparse Mask Optimization

Pre-trained models have been a foundational approach in speech recognition, albeit with associated additional costs. In this study, we propose a regularization technique that facilitates the training of visual and audio-visual speech recognition models (VSR and AVSR) from scratch. This approach, abbreviated as \textbf{MSRS} (Multimodal Speech Recognition from Scratch), introduces a sparse regularization that rapidly learns sparse structures within the dense model at the very beginning of training, which receives healthier gradient flow than the dense equivalent. Once the sparse mask stabilizes, our method allows transitioning to a dense model or keeping a sparse model by updating non-zero values. MSRS achieves competitive results in VSR and AVSR with 21.1% and 0.9% WER on the LRS3 benchmark, while reducing training time by at least 2x. We explore other sparse approaches and show that only MSRS enables training from scratch by implicitly masking the weights affected by vanishing gradients.

cs.CV↗

arXiv2024

CourseGPT-zh: an Educational Large Language Model Based on Knowledge Distillation Incorporating Prompt Optimization

Large language models (LLMs) have demonstrated astonishing capabilities in natural language processing (NLP) tasks, sparking interest in their application to professional domains with higher specialized requirements. However, restricted access to closed-source LLMs via APIs and the difficulty in collecting massive high-quality datasets pose obstacles to the development of large language models in education fields of various courses. Given these challenges, we propose CourseGPT-zh, a course-oriented education LLM that supports customization and low-cost deployment. To address the comprehensiveness and diversity requirements of course-specific corpora, we design a high-quality question-answering corpus distillation framework incorporating prompt optimization, which effectively mines textbook knowledge and enhances its diversity. Moreover, considering the alignment of LLM responses with user needs, a novel method for discrete prompt optimization based on LLM-as-Judge is introduced. During optimization, this framework leverages the LLM's ability to reflect on and exploit error feedback and patterns, allowing for prompts that meet user needs and preferences while saving response length. Lastly, we obtain CourseGPT-zh based on the open-source LLM using parameter-efficient fine-tuning. Experimental results show that our discrete prompt optimization framework effectively improves the response quality of ChatGPT, and CourseGPT-zh exhibits strong professional capabilities in specialized knowledge question-answering, significantly outperforming comparable open-source models.

cs.CL↗

arXiv2024

FFN-SkipLLM: A Hidden Gem for Autoregressive Decoding with Adaptive Feed Forward Skipping

Autoregressive Large Language Models (e.g., LLaMa, GPTs) are omnipresent achieving remarkable success in language understanding and generation. However, such impressive capability typically comes with a substantial model size, which presents significant challenges for autoregressive token-by-token generation. To mitigate computation overload incurred during generation, several early-exit and layer-dropping strategies have been proposed. Despite some promising success due to the redundancy across LLMs layers on metrics like Rough-L/BLUE, our careful knowledge-intensive evaluation unveils issues such as generation collapse, hallucination of wrong facts, and noticeable performance drop even at the trivial exit ratio of 10-15% of layers. We attribute these errors primarily to ineffective handling of the KV cache through state copying during early-exit. In this work, we observed the saturation of computationally expensive feed-forward blocks of LLM layers and proposed FFN-SkipLLM, which is a novel fine-grained skip strategy of autoregressive LLMs. More specifically, FFN-SkipLLM is an input-adaptive feed-forward skipping strategy that can skip 25-30% of FFN blocks of LLMs with marginal change in performance on knowledge-intensive generation tasks without any requirement to handle KV cache. Our extensive experiments and ablation across benchmarks like MT-Bench, Factoid-QA, and variable-length text summarization illustrate how our simple and ease-at-use method can facilitate faster autoregressive decoding.

cs.CL↗

arXiv2024

You Can Have Better Graph Neural Networks by Not Training Weights at All: Finding Untrained GNNs Tickets

Recent works have impressively demonstrated that there exists a subnetwork in randomly initialized convolutional neural networks (CNNs) that can match the performance of the fully trained dense networks at initialization, without any optimization of the weights of the network (i.e., untrained networks). However, the presence of such untrained subnetworks in graph neural networks (GNNs) still remains mysterious. In this paper we carry out the first-of-its-kind exploration of discovering matching untrained GNNs. With sparsity as the core tool, we can find \textit{untrained sparse subnetworks} at the initialization, that can match the performance of \textit{fully trained dense} GNNs. Besides this already encouraging finding of comparable performance, we show that the found untrained subnetworks can substantially mitigate the GNN over-smoothing problem, hence becoming a powerful tool to enable deeper GNNs without bells and whistles. We also observe that such sparse untrained subnetworks have appealing performance in out-of-distribution detection and robustness of input perturbations. We evaluate our method across widely-used GNN architectures on various popular datasets including the Open Graph Benchmark (OGB).

cs.LG↗

arXiv2024

Enhancing Adversarial Training via Reweighting Optimization Trajectory

Despite the fact that adversarial training has become the de facto method for improving the robustness of deep neural networks, it is well-known that vanilla adversarial training suffers from daunting robust overfitting, resulting in unsatisfactory robust generalization. A number of approaches have been proposed to address these drawbacks such as extra regularization, adversarial weights perturbation, and training with more data over the last few years. However, the robust generalization improvement is yet far from satisfactory. In this paper, we approach this challenge with a brand new perspective -- refining historical optimization trajectories. We propose a new method named \textbf{Weighted Optimization Trajectories (WOT)} that leverages the optimization trajectories of adversarial training in time. We have conducted extensive experiments to demonstrate the effectiveness of WOT under various state-of-the-art adversarial attacks. Our results show that WOT integrates seamlessly with the existing adversarial training methods and consistently overcomes the robust overfitting issue, resulting in better adversarial robustness. For example, WOT boosts the robust accuracy of AT-PGD under AA-$L_{\infty}$ attack by 1.53\% $\sim$ 6.11\% and meanwhile increases the clean accuracy by 0.55\%$\sim$5.47\% across SVHN, CIFAR-10, CIFAR-100, and Tiny-ImageNet datasets.

cs.LG↗

arXiv2023

REST: Enhancing Group Robustness in DNNs through Reweighted Sparse Training

The deep neural network (DNN) has been proven effective in various domains. However, they often struggle to perform well on certain minority groups during inference, despite showing strong performance on the majority of data groups. This is because over-parameterized models learned \textit{bias attributes} from a large number of \textit{bias-aligned} training samples. These bias attributes are strongly spuriously correlated with the target variable, causing the models to be biased towards spurious correlations (i.e., \textit{bias-conflicting}). To tackle this issue, we propose a novel \textbf{re}weighted \textbf{s}parse \textbf{t}raining framework, dubbed as \textit{\textbf{REST}}, which aims to enhance the performance of biased data while improving computation and memory efficiency. Our proposed REST framework has been experimentally validated on three datasets, demonstrating its effectiveness in exploring unbiased subnetworks. We found that REST reduces the reliance on spuriously correlated features, leading to better performance across a wider range of data groups with fewer training and inference resources. We highlight that the \textit{REST} framework represents a promising approach for improving the performance of DNNs on biased data, while simultaneously improving computation and memory efficiency. By reducing the reliance on spurious correlations, REST has the potential to enhance the robustness of DNNs and improve their generalization capabilities. Code is released at \url{https://github.com/zhao1402072392/REST}

cs.LG↗

arXiv2023

A Structural-Clustering Based Active Learning for Graph Neural Networks

In active learning for graph-structured data, Graph Neural Networks (GNNs) have shown effectiveness. However, a common challenge in these applications is the underutilization of crucial structural information. To address this problem, we propose the Structural-Clustering PageRank method for improved Active learning (SPA) specifically designed for graph-structured data. SPA integrates community detection using the SCAN algorithm with the PageRank scoring method for efficient and informative sample selection. SPA prioritizes nodes that are not only informative but also central in structure. Through extensive experiments, SPA demonstrates higher accuracy and macro-F1 score over existing methods across different annotation budgets and achieves significant reductions in query time. In addition, the proposed method only adds two hyperparameters, $ε$ and $μ$ in the algorithm to finely tune the balance between structural learning and node selection. This simplicity is a key advantage in active learning scenarios, where extensive hyperparameter tuning is often impractical.

cs.LG↗

arXiv2023

Dynamic Sparsity Is Channel-Level Sparsity Learner

Sparse training has received an upsurging interest in machine learning due to its tantalizing saving potential for the entire training process as well as inference. Dynamic sparse training (DST), as a leading sparse training approach, can train deep neural networks at high sparsity from scratch to match the performance of their dense counterparts. However, most if not all DST prior arts demonstrate their effectiveness on unstructured sparsity with highly irregular sparse patterns, which receives limited support in common hardware. This limitation hinders the usage of DST in practice. In this paper, we propose Channel-aware dynamic sparse (Chase), which for the first time seamlessly translates the promise of unstructured dynamic sparsity to GPU-friendly channel-level sparsity (not fine-grained N:M or group sparsity) during one end-to-end training process, without any ad-hoc operations. The resulting small sparse networks can be directly accelerated by commodity hardware, without using any particularly sparsity-aware hardware accelerators. This appealing outcome is partially motivated by a hidden phenomenon of dynamic sparsity: off-the-shelf unstructured DST implicitly involves biased parameter reallocation across channels, with a large fraction of channels (up to 60%) being sparser than others. By progressively identifying and removing these channels during training, our approach translates unstructured sparsity to channel-wise sparsity. Our experimental results demonstrate that Chase achieves 1.7 X inference throughput speedup on common GPU devices without compromising accuracy with ResNet-50 on ImageNet. We release our codes in https://github.com/luuyin/chase.

cs.LG↗

arXiv2023

WIMPs in Dilatonic Einstein Gauss-Bonnet Cosmology

We use the Weakly Interacting Massive Particle (WIMP) thermal decoupling scenario to probe Cosmologies in dilatonic Einstein Gauss-Bonnet (dEGB) gravity, where the Gauss-Bonnet term is non-minimally coupled to a scalar field with vanishing potential. We put constraints on the model parameters when the ensuing modified cosmological scenario drives the WIMP annihilation cross section beyond the present bounds from DM indirect detection searches. In our analysis we assumed WIMPs that annihilate to Standard Model particles through an s-wave process. For the class of solutions that comply with WIMP indirect detection bounds, we find that dEGB typically plays a mitigating role on the scalar field dynamics at high temperature, slowing down the speed of its evolution and reducing the enhancement of the Hubble constant compared to its standard value. For such solutions, we observe that the corresponding boundary conditions at high temperature correspond asymptotically to a vanishing deceleration parameter q, so that the effect of dEGB is to add an accelerating term that exactly cancels the deceleration predicted by General Relativity. The bounds from WIMP indirect detection are nicely complementary to late-time constraints from compact binary mergers. This suggest that it could be interesting to use other Early Cosmology processes to probe the dEGB scenario.

hep-ph↗

arXiv2023

Inevitable manifestation of wiggles in the expansion of the late Universe

Using the fact that the comoving angular diameter distance to the last scattering surface is strictly constrained almost model independently, we show that, for any model agreeing with the standard $Λ$CDM model on its background dynamics at $z\sim0$ and size of the comoving sound horizon at last scattering, the deviations of the Hubble radius from the one of the standard $Λ$CDM model must be a member of the set of admissible wavelets. The family of models characterized by this framework also offers nontrivial oscillatory behaviours in various functions that define the kinematics of the Universe, even when the wavelets themselves are very simple. We also discuss the consequences of attributing these kinematics to, first, dark energy, and second, varying gravitational coupling strength. Utilizing some simplest wavelets, we demonstrate the competence of this framework in describing the baryon acoustic oscillation (BAO) data without any modifications to the agreement with cosmic microwave background measurements. This framework also provides a natural explanation for the bumps found in nonparametric observational reconstructions of the Hubble parameter and dark energy density as compensations of the dips suggested by some BAO data, and questions the physical reality of their existence. We note that utilizing this framework on top of the models that agree with both the cosmic microwave background and local $H_0$ measurements but are held back by BAO data, one may resurrect these models through the wiggly nature of wavelets that can naturally accommodate the BAO data. Finally, we also suggest narrowing the plausible set of admissible wavelets to further improve our framework by imposing conditions from expected kinematics of a viable cosmological model or first principle fundamental physics such as energy conditions.

astro-ph.CO↗

arXiv2023

Cosmic-Dawn 21-cm Signal from Dynamical Dark Energy

The 21-cm signal is the most important measurement for us to understand physics during cosmic dawn. It is the key for us to understand the expansion history of the Universe and the nature of dark energy. In this paper, we focused on the characteristic 21-cm power spectrum of a special dynamic dark energy - the Interacting Chevallier-Polarski-Linder (ICPL) model - and compared it with those of the $Λ$CDM and CPL models. From the expected noise of HERA, we found more precise experiments in the future can detect the features of interacting dark energy in the 21-cm power spectra. By studying the brightness temperature, we found the ICPL model is closer to the observation of EDGES compared to the $Λ$CDM, thus alleviating the tension between theory and experiments.

astro-ph.CO↗