arXiv subjects

Shiwei Liu

Publications and source records attributed to Shiwei Liu.

107 records · Page 6Linked to original sources

arXiv2022

Superposing Many Tickets into One: A Performance Booster for Sparse Neural Network Training

Recent works on sparse neural network training (sparse training) have shown that a compelling trade-off between performance and efficiency can be achieved by training intrinsically sparse neural networks from scratch. Existing sparse training methods usually strive to find the best sparse subnetwork possible in one single run, without involving any expensive dense or pre-training steps. For instance, dynamic sparse training (DST), is capable of reaching a competitive performance of dense training by iteratively evolving the sparse topology during the course of training. In this paper, we argue that it is better to allocate the limited resources to create multiple low-loss sparse subnetworks and superpose them into a stronger one, instead of allocating all resources entirely to find an individual subnetwork. To achieve this, two desiderata are required: (1) efficiently producing many low-loss subnetworks, the so-called cheap tickets, within one training process limited to the standard training time used in dense training; (2) effectively superposing these cheap tickets into one stronger subnetwork. To corroborate our conjecture, we present a novel sparse training approach, termed Sup-tickets, which can satisfy the above two desiderata concurrently in a single sparse-to-sparse training process. Across various modern architectures on CIFAR-10/100 and ImageNet, we show that Sup-tickets integrates seamlessly with the existing sparse training methods and demonstrates consistent performance improvement.

cs.LG↗

arXiv2022

Don't Be So Dense: Sparse-to-Sparse GAN Training Without Sacrificing Performance

Generative adversarial networks (GANs) have received an upsurging interest since being proposed due to the high quality of the generated data. While achieving increasingly impressive results, the resource demands associated with the large model size hinders the usage of GANs in resource-limited scenarios. For inference, the existing model compression techniques can reduce the model complexity with comparable performance. However, the training efficiency of GANs has less been explored due to the fragile training process of GANs. In this paper, we, for the first time, explore the possibility of directly training sparse GAN from scratch without involving any dense or pre-training steps. Even more unconventionally, our proposed method enables directly training sparse unbalanced GANs with an extremely sparse generator from scratch. Instead of training full GANs, we start with sparse GANs and dynamically explore the parameter space spanned over the generator throughout training. Such a sparse-to-sparse training procedure enhances the capacity of the highly sparse generator progressively while sticking to a fixed small parameter budget with appealing training and inference efficiency gains. Extensive experiments with modern GAN architectures validate the effectiveness of our method. Our sparsified GANs, trained from scratch in one single run, are able to outperform the ones learned by expensive iterative pruning and re-training. Perhaps most importantly, we find instead of inheriting parameters from expensive pre-trained GANs, directly training sparse GANs from scratch can be a much more efficient solution. For example, only training with a 80% sparse generator and a 70% sparse discriminator, our method can achieve even better performance than the dense BigGAN.

cs.CV↗

arXiv2022

Sparse Training via Boosting Pruning Plasticity with Neuroregeneration

Works on lottery ticket hypothesis (LTH) and single-shot network pruning (SNIP) have raised a lot of attention currently on post-training pruning (iterative magnitude pruning), and before-training pruning (pruning at initialization). The former method suffers from an extremely large computation cost and the latter usually struggles with insufficient performance. In comparison, during-training pruning, a class of pruning methods that simultaneously enjoys the training/inference efficiency and the comparable performance, temporarily, has been less explored. To better understand during-training pruning, we quantitatively study the effect of pruning throughout training from the perspective of pruning plasticity (the ability of the pruned networks to recover the original performance). Pruning plasticity can help explain several other empirical observations about neural network pruning in literature. We further find that pruning plasticity can be substantially improved by injecting a brain-inspired mechanism called neuroregeneration, i.e., to regenerate the same number of connections as pruned. We design a novel gradual magnitude pruning (GMP) method, named gradual pruning with zero-cost neuroregeneration (\textbf{GraNet}), that advances state of the art. Perhaps most impressively, its sparse-to-sparse version for the first time boosts the sparse-to-sparse training performance over various dense-to-sparse methods with ResNet-50 on ImageNet without extending the training time. We release all codes in https://github.com/Shiweiliuiiiiiii/GraNet.

cs.LG↗

arXiv2022

The Unreasonable Effectiveness of Random Pruning: Return of the Most Naive Baseline for Sparse Training

Random pruning is arguably the most naive way to attain sparsity in neural networks, but has been deemed uncompetitive by either post-training pruning or sparse training. In this paper, we focus on sparse training and highlight a perhaps counter-intuitive finding, that random pruning at initialization can be quite powerful for the sparse training of modern neural networks. Without any delicate pruning criteria or carefully pursued sparsity structures, we empirically demonstrate that sparsely training a randomly pruned network from scratch can match the performance of its dense equivalent. There are two key factors that contribute to this revival: (i) the network sizes matter: as the original dense networks grow wider and deeper, the performance of training a randomly pruned sparse network will quickly grow to matching that of its dense equivalent, even at high sparsity ratios; (ii) appropriate layer-wise sparsity ratios can be pre-chosen for sparse training, which shows to be another important performance booster. Simple as it looks, a randomly pruned subnetwork of Wide ResNet-50 can be sparsely trained to outperforming a dense Wide ResNet-50, on ImageNet. We also observed such randomly pruned networks outperform dense counterparts in other favorable aspects, such as out-of-distribution detection, uncertainty estimation, and adversarial robustness. Overall, our results strongly suggest there is larger-than-expected room for sparse training at scale, and the benefits of sparsity might be more universal beyond carefully designed pruning. Our source code can be found at https://github.com/VITA-Group/Random_Pruning.

cs.LG↗

arXiv2022

Achieving Personalized Federated Learning with Sparse Local Models

Federated learning (FL) is vulnerable to heterogeneously distributed data, since a common global model in FL may not adapt to the heterogeneous data distribution of each user. To counter this issue, personalized FL (PFL) was proposed to produce dedicated local models for each individual user. However, PFL is far from its maturity, because existing PFL solutions either demonstrate unsatisfactory generalization towards different model architectures or cost enormous extra computation and memory. In this work, we propose federated learning with personalized sparse mask (FedSpa), a novel PFL scheme that employs personalized sparse masks to customize sparse local models on the edge. Instead of training an intact (or dense) PFL model, FedSpa only maintains a fixed number of active parameters throughout training (aka sparse-to-sparse training), which enables users' models to achieve personalization with cheap communication, computation, and memory cost. We theoretically show that the iterates obtained by FedSpa converge to the local minimizer of the formulated SPFL problem at rate of $\mathcal{O}(\frac{1}{\sqrt{T}})$. Comprehensive experiments demonstrate that FedSpa significantly saves communication and computation costs, while simultaneously achieves higher model accuracy and faster convergence speed against several state-of-the-art PFL methods.

cs.LG↗

arXiv2021

Hierarchical Semantic Segmentation using Psychometric Learning

Assigning meaning to parts of image data is the goal of semantic image segmentation. Machine learning methods, specifically supervised learning is commonly used in a variety of tasks formulated as semantic segmentation. One of the major challenges in the supervised learning approaches is expressing and collecting the rich knowledge that experts have with respect to the meaning present in the image data. Towards this, typically a fixed set of labels is specified and experts are tasked with annotating the pixels, patches or segments in the images with the given labels. In general, however, the set of classes does not fully capture the rich semantic information present in the images. For example, in medical imaging such as histology images, the different parts of cells could be grouped and sub-grouped based on the expertise of the pathologist. To achieve such a precise semantic representation of the concepts in the image, we need access to the full depth of knowledge of the annotator. In this work, we develop a novel approach to collect segmentation annotations from experts based on psychometric testing. Our method consists of the psychometric testing procedure, active query selection, query enhancement, and a deep metric learning model to achieve a patch-level image embedding that allows for semantic segmentation of images. We show the merits of our method with evaluation on the synthetically generated image, aerial image and histology image.

cs.CV↗

arXiv2021

Physical meaning of the deviation scale under arbitrary turbulence strengths of optical orbital angular momentum

The recently so-called deviation scale [C. M. Mabena et al., Phys. Rev. A 99, 013828 (2019)] bridges the connection between the result of the infinitesimal propagation equation (IPE) prediction and that of the single phase screen (SPS) approximation. Thanks to the multiple phase screen (MPS) approach, in this paper we elaborate the physical meaning of the deviation scale: the spatial accumulation of slight intensity modulation of incident orbital angular momentum (OAM) carrying beam splits the original vortex into multiple individual vortices with a topological charge (TC) of +1 and re-generates the vortex-antivortex pairs with a TC of +1 and with a TC of -1, leading to a significant deviation between these two different results only when the disruption of this compound effect on the phase distribution of the incident OAM-carrying beam becomes more significant. Other than that, we also show that the appearance of the deviation scale cannot be predicted only by the Rytov variance, which can be predicted through the vortex-splitting ratio of the received optical field alone or with the help of the normalized propagation distance.

physics.optics↗

arXiv2021

Selfish Sparse RNN Training

Sparse neural networks have been widely applied to reduce the computational demands of training and deploying over-parameterized deep neural networks. For inference acceleration, methods that discover a sparse network from a pre-trained dense network (dense-to-sparse training) work effectively. Recently, dynamic sparse training (DST) has been proposed to train sparse neural networks without pre-training a dense model (sparse-to-sparse training), so that the training process can also be accelerated. However, previous sparse-to-sparse methods mainly focus on Multilayer Perceptron Networks (MLPs) and Convolutional Neural Networks (CNNs), failing to match the performance of dense-to-sparse methods in the Recurrent Neural Networks (RNNs) setting. In this paper, we propose an approach to train intrinsically sparse RNNs with a fixed parameter count in one single run, without compromising performance. During training, we allow RNN layers to have a non-uniform redistribution across cell gates for better regularization. Further, we propose SNT-ASGD, a novel variant of the averaged stochastic gradient optimizer, which significantly improves the performance of all sparse training methods for RNNs. Using these strategies, we achieve state-of-the-art sparse training results, better than the dense-to-sparse methods, with various types of RNNs on Penn TreeBank and Wikitext-2 datasets. Our codes are available at https://github.com/Shiweiliuiiiiiii/Selfish-RNN.

cs.LG↗

arXiv2021

Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training

In this paper, we introduce a new perspective on training deep neural networks capable of state-of-the-art performance without the need for the expensive over-parameterization by proposing the concept of In-Time Over-Parameterization (ITOP) in sparse training. By starting from a random sparse network and continuously exploring sparse connectivities during training, we can perform an Over-Parameterization in the space-time manifold, closing the gap in the expressibility between sparse training and dense training. We further use ITOP to understand the underlying mechanism of Dynamic Sparse Training (DST) and indicate that the benefits of DST come from its ability to consider across time all possible parameters when searching for the optimal sparse connectivity. As long as there are sufficient parameters that have been reliably explored during training, DST can outperform the dense neural network by a large margin. We present a series of experiments to support our conjecture and achieve the state-of-the-art sparse training performance with ResNet-50 on ImageNet. More impressively, our method achieves dominant performance over the overparameterization-based sparse methods at extreme sparsity levels. When trained on CIFAR-100, our method can match the performance of the dense model even at an extreme sparsity (98%). Code can be found https://github.com/Shiweiliuiiiiiii/In-Time-Over-Parameterization.

cs.LG↗

arXiv2021

Phase-sensitive nonclassical properties in quantum metrology with a displaced squeezed vacuum state

We predict that the phase-dependent error distribution of locally unentangled quantum states directly affects quantum parameter estimation accuracy. Therefore, we employ the displaced squeezed vacuum (DSV) state as a probe state and investigate an interesting question of the phase-sensitive nonclassical properties in DSV's metrology. We found that the accuracy limit of parameter estimation is a function of the phase-sensitive parameter $ϕ-θ/2$ with a period $π$. We show that when $ϕ-θ/2$ $\in \left[ kπ/2,3kπ/4\right) \left( k\in \mathbb{Z}\right)$, we can obtain the accuracy of parameter estimation approaching the ultimate quantum limit through using the DSV state with the larger displacement and squeezing strength, whereas $ϕ-θ/2$ $\in \left(3kπ/4,kπ\right] \left( k\in \mathbb{Z}\right) $, the optimal estimation accuracy can be acquired only when the DSV state degenerates to a squeezed-vacuum state.

quant-ph↗

arXiv2021

Mitigating the effect of atmospheric turbulence on orbital angular momentum-based quantum key distribution using real-time adaptive optics with phase unwrapping

Quantum key distribution (QKD) employed orbital angular momentum (OAM) for high-dimensional encoding enhances the system security and information capacity between two communication parties. However, such advantages significantly degrade because of the fragility of OAM states in atmospheric turbulence. Unlike previous researches, we first investigate the performance degradation of OAM-based QKD by infinitely long phase screen (ILPS), which offers a feasible way to study how adaptive optics (AO) dynamically corrects the turbulence-induced aberrations in real time. Secondly, considering the failure of AO while encountering phase cuts, we evaluate the quality enhancement of OAM-based QKD under moderate turbulence strengths by AO after implementing the wrapped cuts elimination. Finally, we simulate that, with more realistic considerations, real-time AO can still mitigate the impact of atmospheric turbulence on OAM-based QKD even in the large wind velocity regime.

quant-ph↗

arXiv2021

Deuterium-tritium fusion process in strong laser fields: Semiclassical simulation

In this paper, we investigate the deuterium-tritium (DT) fusion process in the presence of strong laser fields with a semiclassical (SC) method. In this model, two nuclei with a given incident kinetic energy that closely approach each other are simulated by tracing the classical Newtonian trajectories in the combined Coulomb repulsive potentials and laser fields. At the nearest position or classical turning point, quantum tunneling through the Coulomb barrier emerges, and its penetrability is estimated with the Wentzel-Kramers-Brillouin formula. Nuclear fusion occurs after the tunneling, and the total fusion cross section takes the Gamow form. We find that the tunneling penetrability can be enhanced dramatically because the nuclei can closely approach each other due to the quiver motion of the charged nuclei driven by the intense laser fields. We then calculate the DT fusion section for a wide range of laser parameters according to various incident nuclei kinetic energies and obtain the phase diagrams for the enhanced DT fusion. We compare our SC results with the quantum results of the Kramers-Henneberger approximation and the Volkov state approximation.

nucl-th↗

arXiv2021

Sparse evolutionary Deep Learning with over one million artificial neurons on commodity hardware

Artificial Neural Networks (ANNs) have emerged as hot topics in the research community. Despite the success of ANNs, it is challenging to train and deploy modern ANNs on commodity hardware due to the ever-increasing model size and the unprecedented growth in the data volumes. Particularly for microarray data, the very-high dimensionality and the small number of samples make it difficult for machine learning techniques to handle. Furthermore, specialized hardware such as Graphics Processing Unit (GPU) is expensive. Sparse neural networks are the leading approaches to address these challenges. However, off-the-shelf sparsity inducing techniques either operate from a pre-trained model or enforce the sparse structure via binary masks. The training efficiency of sparse neural networks cannot be obtained practically. In this paper, we introduce a technique allowing us to train truly sparse neural networks with fixed parameter count throughout training. Our experimental results demonstrate that our method can be applied directly to handle high dimensional data, while achieving higher accuracy than the traditional two phases approaches. Moreover, we have been able to create truly sparse MultiLayer Perceptrons (MLPs) models with over one million neurons and to train them on a typical laptop without GPU, this being way beyond what is possible with any state-of-the-art techniques.

cs.NE↗

arXiv2020

Topological Insights into Sparse Neural Networks

Sparse neural networks are effective approaches to reduce the resource requirements for the deployment of deep neural networks. Recently, the concept of adaptive sparse connectivity, has emerged to allow training sparse neural networks from scratch by optimizing the sparse structure during training. However, comparing different sparse topologies and determining how sparse topologies evolve during training, especially for the situation in which the sparse structure optimization is involved, remain as challenging open questions. This comparison becomes increasingly complex as the number of possible topological comparisons increases exponentially with the size of networks. In this work, we introduce an approach to understand and compare sparse neural network topologies from the perspective of graph theory. We first propose Neural Network Sparse Topology Distance (NNSTD) to measure the distance between different sparse neural networks. Further, we demonstrate that sparse neural networks can outperform over-parameterized models in terms of performance, even without any further structure optimization. To the end, we also show that adaptive sparse connectivity can always unveil a plenitude of sparse sub-networks with very different topologies which outperform the dense model, by quantifying and comparing their topological evolutionary processes. The latter findings complement the Lottery Ticket Hypothesis by showing that there is a much more efficient and robust way to find "winning tickets". Altogether, our results start enabling a better theoretical understanding of sparse neural networks, and demonstrate the utility of using graph theory to analyze them.

cs.LG↗

arXiv2020

Ellipticity-dependent sequential over-barrier ionization of cold rubidium

We perform high-resolution measurements of momentum distribution on Rb$^{n+}$ recoil ions up to charge state $n=4$, where laser-cooled rubidium atoms are ionized by femtosecond elliptically polarized lasers with the pulse duration of 35 fs and the intensity of 3.3$\times$10$^{15}$ W/cm$^2$ in the over-barrier ionization (OBI) regime. The momentum distributions of the recoil ions are found to exhibit multi-band structures as the ellipticity varies from the linear to circular polarizations. The origin of these band structures can be explained quantitatively by the classical OBI model and dedicated classical trajectory Monte Carlo simulations with Heisenberg potential. Specifically, with back analysis of the classical trajectories, we reveal the ionization time and the OBI geometry of the sequentially released electrons, disentangling the mechanisms behind the tilted angle of the band structures. These results indicate that the classical treatment can describe the strong-field multiple ionization processes of alkali atoms.

physics.atom-ph↗

arXiv2019

On improving deep learning generalization with adaptive sparse connectivity

Large neural networks are very successful in various tasks. However, with limited data, the generalization capabilities of deep neural networks are also very limited. In this paper, we empirically start showing that intrinsically sparse neural networks with adaptive sparse connectivity, which by design have a strict parameter budget during the training phase, have better generalization capabilities than their fully-connected counterparts. Besides this, we propose a new technique to train these sparse models by combining the Sparse Evolutionary Training (SET) procedure with neurons pruning. Operated on MultiLayer Perceptron (MLP) and tested on 15 datasets, our proposed technique zeros out around 50% of the hidden neurons during training, while having a linear number of parameters to optimize with respect to the number of neurons. The results show a competitive classification and generalization performance.

cs.NE↗

arXiv2019

Intrinsically Sparse Long Short-Term Memory Networks

Long Short-Term Memory (LSTM) has achieved state-of-the-art performances on a wide range of tasks. Its outstanding performance is guaranteed by the long-term memory ability which matches the sequential data perfectly and the gating structure controlling the information flow. However, LSTMs are prone to be memory-bandwidth limited in realistic applications and need an unbearable period of training and inference time as the model size is ever-increasing. To tackle this problem, various efficient model compression methods have been proposed. Most of them need a big and expensive pre-trained model which is a nightmare for resource-limited devices where the memory budget is strictly limited. To remedy this situation, in this paper, we incorporate the Sparse Evolutionary Training (SET) procedure into LSTM, proposing a novel model dubbed SET-LSTM. Rather than starting with a fully-connected architecture, SET-LSTM has a sparse topology and dramatically fewer parameters in both phases, training and inference. Considering the specific architecture of LSTMs, we replace the LSTM cells and embedding layers with sparse structures and further on, use an evolutionary strategy to adapt the sparse connectivity to the data. Additionally, we find that SET-LSTM can provide many different good combinations of sparse connectivity to substitute the overparameterized optimization problem of dense neural networks. Evaluated on four sentiment analysis classification datasets, the results demonstrate that our proposed model is able to achieve usually better performance than its fully connected counterpart while having less than 4\% of its parameters.

cs.NE↗