Many existing Neural Network pruning approaches rely on either retraining or inducing a strong bias in order to converge to a sparse solution throughout training. A third paradigm, 'compression-aware' training, aims to obtain state-of-the-art dense models that are robust to a wide range of compression ratios using a single dense training run while also avoiding retraining. We propose a framework centered around a versatile family of norm constraints and the Stochastic Frank-Wolfe (SFW) algorithm that encourage convergence to well-performing solutions while inducing robustness towards convolutional filter pruning and low-rank matrix decomposition. Our method is able to outperform existing compression-aware approaches and, in the case of low-rank matrix decomposition, it also requires significantly less computational resources than approaches based on nuclear-norm regularization. Our findings indicate that dynamically adjusting the learning rate of SFW, as suggested by Pokutta et al. (2020), is crucial for convergence and robustness of SFW-trained models and we establish a theoretical foundation for that practice.
相關內容
The letter focuses on Haptic Glove (HG) based control of a Robotic Hand (RH) executing in-hand manipulation of certain objects of interest. The high dimensional motion signals in HG and RH possess intrinsic variability of kinematics resulting in difficulty to establish a direct mapping of the motion signals from HG onto the RH. An estimation mechanism is proposed to quantify the motion signal acquired from the human controller in relation to the intended goal pose of the object being held by the robotic hand. A control algorithm is presented to transform the synthesized intent at the RH and allow relocation of the object to the expected goal pose. The lag in synthesis of the intent in the presence of communication delay leads to a requirement of predicting the estimated intent. We leverage an attention-based convolutional neural network encoder to predict the trajectory of intent for a certain lookahead to compensate for the delays. The proposed methodology is evaluated across objects of different shapes, mass, and materials. We present a comparative performance of the estimation and prediction mechanisms on 5G-driven real-world robotic setup against benchmark methodologies. The test-MSE in prediction of human intent is reported to yield ~ 97.3 -98.7% improvement of accuracy in comparison to LSTM-based benchmark
Deciding what combination of operators to use across the Edge AI tiers to achieve specific latency and model performance requirements is an open question for MLOps engineers. This study aims to empirically assess the accuracy vs inference time trade-off of different black-box Edge AI deployment strategies, i.e., combinations of deployment operators and deployment tiers. In this paper, we conduct inference experiments involving 3 deployment operators (i.e., Partitioning, Quantization, Early Exit), 3 deployment tiers (i.e., Mobile, Edge, Cloud) and their combinations on four widely used Computer-Vision models to investigate the optimal strategies from the point of view of MLOps developers. Our findings suggest that Edge deployment using the hybrid Quantization + Early Exit operator could be preferred over non-hybrid operators (Quantization/Early Exit on Edge, Partition on Mobile-Edge) when faster latency is a concern at medium accuracy loss. However, when minimizing accuracy loss is a concern, MLOps engineers should prefer using only a Quantization operator on edge at a latency reduction or increase, respectively over the Early Exit/Partition (on edge/mobile-edge) and Quantized Early Exit (on edge) operators. In scenarios constrained by Mobile CPU/RAM resources, a preference for Partitioning across mobile and edge tiers is observed over mobile deployment. For models with smaller input data samples (such as FCN), a network-constrained cloud deployment can also be a better alternative than Mobile/Edge deployment and Partitioning strategies. For models with large input data samples (ResNet, ResNext, DUC), an edge tier having higher network/computational capabilities than Cloud/Mobile can be a more viable option than Partitioning and Mobile/Cloud deployment strategies.
We interact with the world with our hands and see it through our own (egocentric) perspective. A holistic 3D understanding of such interactions from egocentric views is important for tasks in robotics, AR/VR, action recognition and motion generation. Accurately reconstructing such interactions in 3D is challenging due to heavy occlusion, viewpoint bias, camera distortion, and motion blur from the head movement. To this end, we designed the HANDS23 challenge based on the AssemblyHands and ARCTIC datasets with carefully designed training and testing splits. Based on the results of the top submitted methods and more recent baselines on the leaderboards, we perform a thorough analysis on 3D hand(-object) reconstruction tasks. Our analysis demonstrates the effectiveness of addressing distortion specific to egocentric cameras, adopting high-capacity transformers to learn complex hand-object interactions, and fusing predictions from different views. Our study further reveals challenging scenarios intractable with state-of-the-art methods, such as fast hand motion, object reconstruction from narrow egocentric views, and close contact between two hands and objects. Our efforts will enrich the community's knowledge foundation and facilitate future hand studies on egocentric hand-object interactions.
Edge AI has been recently proposed to facilitate the training and deployment of Deep Neural Network (DNN) models in proximity to the sources of data. To enable the training of large models on resource-constraint edge devices and protect data privacy, parallel split learning is becoming a practical and popular approach. However, current parallel split learning neglects the resource heterogeneity of edge devices, which may lead to the straggler issue. In this paper, we propose EdgeSplit, a novel parallel split learning framework to better accelerate distributed model training on heterogeneous and resource-constraint edge devices. EdgeSplit enhances the efficiency of model training on less powerful edge devices by adaptively segmenting the model into varying depths. Our approach focuses on reducing total training time by formulating and solving a task scheduling problem, which determines the most efficient model partition points and bandwidth allocation for each device. We employ a straightforward yet effective alternating algorithm for this purpose. Comprehensive tests conducted with a range of DNN models and datasets demonstrate that EdgeSplit not only facilitates the training of large models on resource-restricted edge devices but also surpasses existing baselines in performance.
This paper aims to clearly distinguish between Stochastic Gradient Descent with Momentum (SGDM) and Adam in terms of their convergence rates. We demonstrate that Adam achieves a faster convergence compared to SGDM under the condition of non-uniformly bounded smoothness. Our findings reveal that: (1) in deterministic environments, Adam can attain the known lower bound for the convergence rate of deterministic first-order optimizers, whereas the convergence rate of Gradient Descent with Momentum (GDM) has higher order dependence on the initial function value; (2) in stochastic setting, Adam's convergence rate upper bound matches the lower bounds of stochastic first-order optimizers, considering both the initial function value and the final error, whereas there are instances where SGDM fails to converge with any learning rate. These insights distinctly differentiate Adam and SGDM regarding their convergence rates. Additionally, by introducing a novel stopping-time based technique, we further prove that if we consider the minimum gradient norm during iterations, the corresponding convergence rate can match the lower bounds across all problem hyperparameters. The technique can also help proving that Adam with a specific hyperparameter scheduler is parameter-agnostic, which hence can be of independent interest.
Traditional pruning methods are known to be challenging to work in Large Language Models (LLMs) for Generative AI because of their unaffordable training process and large computational demands. For the first time, we introduce the information entropy of hidden state features into a pruning metric design, namely E-Sparse, to improve the accuracy of N:M sparsity on LLM. E-Sparse employs the information richness to leverage the channel importance, and further incorporates several novel techniques to put it into effect: (1) it introduces information entropy to enhance the significance of parameter weights and input feature norms as a novel pruning metric, and performs N:M sparsity without modifying the remaining weights. (2) it designs global naive shuffle and local block shuffle to quickly optimize the information distribution and adequately cope with the impact of N:M sparsity on LLMs' accuracy. E-Sparse is implemented as a Sparse-GEMM on FasterTransformer and runs on NVIDIA Ampere GPUs. Extensive experiments on the LLaMA family and OPT models show that E-Sparse can significantly speed up the model inference over the dense model (up to 1.53X) and obtain significant memory saving (up to 43.52%), with acceptable accuracy loss.
Sarcasm is a way of verbal irony where someone says the opposite of what they mean, often to ridicule a person, situation, or idea. It is often difficult to detect sarcasm in the dialogue since detecting sarcasm should reflect the context (i.e., dialogue history). In this paper, we introduce a new dataset for the Korean dialogue sarcasm detection task, KoCoSa (Korean Context-aware Sarcasm Detection Dataset), which consists of 12.8K daily Korean dialogues and the labels for this task on the last response. To build the dataset, we propose an efficient sarcasm detection dataset generation pipeline: 1) generating new sarcastic dialogues from source dialogues with large language models, 2) automatic and manual filtering of abnormal and toxic dialogues, and 3) human annotation for the sarcasm detection task. We also provide a simple but effective baseline for the Korean sarcasm detection task trained on our dataset. Experimental results on the dataset show that our baseline system outperforms strong baselines like large language models, such as GPT-3.5, in the Korean sarcasm detection task. We show that the sarcasm detection task relies deeply on the existence of sufficient context. We will release the dataset at //github.com/Yu-billie/KoCoSa_sarcasm_detection.
Existence constraints were defined in the Relational Data Model, but, unfortunately, are not provided by any Relational Database Management System, except for their NOT NULL particular case. Our (Elementary) Mathematical Data Model extended them to function products and introduced their dual non-existence constraints. MatBase, an intelligent data and knowledge base management system prototype based on both these data models, not only provides existence and non-existence constraints, but also automatically generates code for their enforcement. This paper presents and discusses the algorithms used by MatBase to enforce these types of constraints.
This research investigates the antecedents of positive and negative electronic word-of-mouth (eWOM) propensity, as well as the impact of eWOM propensity on the intention to repurchase the product. Two types of eWOM predictors were considered: product related variables and personal factors. The data were collected through an online survey conducted on a sample of 335 Romanian subjects, and the analysis method was Structural Equation Modeling. Our findings show that personal factors - social media usage behavior, marketing mavenism and need to evaluate - are the most important antecedents of the intention to write product reviews and comments online, either positive or negative. From the product related factors, only brand trust influences the propensity to provide eWOM. Furthermore, both positive and negative eWOM intentions are associated with the repurchase intention.
The dominating NLP paradigm of training a strong neural predictor to perform one task on a specific dataset has led to state-of-the-art performance in a variety of applications (eg. sentiment classification, span-prediction based question answering or machine translation). However, it builds upon the assumption that the data distribution is stationary, ie. that the data is sampled from a fixed distribution both at training and test time. This way of training is inconsistent with how we as humans are able to learn from and operate within a constantly changing stream of information. Moreover, it is ill-adapted to real-world use cases where the data distribution is expected to shift over the course of a model's lifetime. The first goal of this thesis is to characterize the different forms this shift can take in the context of natural language processing, and propose benchmarks and evaluation metrics to measure its effect on current deep learning architectures. We then proceed to take steps to mitigate the effect of distributional shift on NLP models. To this end, we develop methods based on parametric reformulations of the distributionally robust optimization framework. Empirically, we demonstrate that these approaches yield more robust models as demonstrated on a selection of realistic problems. In the third and final part of this thesis, we explore ways of efficiently adapting existing models to new domains or tasks. Our contribution to this topic takes inspiration from information geometry to derive a new gradient update rule which alleviate catastrophic forgetting issues during adaptation.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191