Search

Orbital angular momentum (OAM), which is considered as a novel way to achieve high capacity, has been attracted much attention recently. OAM signals emitted by uniform circular array (UCA) are widely regarded to go through the Bessel-form channels. However, the channel gains corresponding to the Bessel-form channels are with low signal-to-noise-ratio (SNR) on OAM-modes and it is difficult to achieve high capacity using all OAM modes. To achieve maximum capacity offered by OAM multiplexing for wireless backhaul communications, in this paper we propose the optimal UCA design, which selects the optimal OAM-modes and radius of receive UCA. We formulate the capacity maximization problem and divide it into two subproblems for obtaining the corresponding optimal UCA design for OAM multiplexing based wireless backhaul communications. In particular, the optimal radius of the receive UCA is firstly derived. Then, we propose an mode selection scheme to choose appropriate OAM-modes for data transmission to maximize the capacity. Extensive simulations obtained validate that the capacity of OAM multiplexing can be significantly increased with our developed scheme.

Orbital angular momentum (OAM) has attracted much attention for radio vortex wireless communications due to the orthogonality among different OAM-modes. To maintain the orthogonality among different OAM modes at the receiver, the strict alignment between transmit and receive antennas is highly demanded. However, it is not practical to guarantee the transceiver alignment in wireless communications. The phase turbulence, resulting from the misaligned transceivers, leads to serious inter-mode interference among different OAM modes and therefore fail for signals detection of multiple OAM modes at the receiver. To achieve practical OAM based wireless communications, in this paper we investigate the radio vortex wireless communications with misaligned transmit and receive antennas. We propose a joint Beamforming and Pre-detection (BePre) scheme, which uses two unitary matrices to convert the channel matrix into the equivalent circulant matrix for keeping the orthogonality among OAM-modes at the receiver. Then, the OAM signals can be detected with the mode-decomposition scheme at the misaligned receiver. Extensive simulations obtained validate and evaluate that our developed joint BePre scheme can efficiently detect the signals of multiple OAM-modes for the misaligned transceiver and can significantly increase the spectrum efficiency.

Due to the potential capacity-boosting for wireless communications, the Radio vOrtex Wireless COMMunication (RowComm) over orthogonal states/modes of Orbital Angular Momentum (OAM) has been paid much attention in recent years. Uniform circular array (UCA), as an efficient and convenient antenna structure, can transmit/receive multiple OAM beams with different OAM-modes simultaneously when the transmitter and receiver are aligned. However, for high-order OAM-modes, the OAM beams are divergent accompanied by severe attenuations. Thus, it is difficult to directly use high-order OAM-modes to achieve high capacity for RowComms. To obtain high capacity potentially offered by OAM-modes, in this paper we transform the singular UCA into the concentric UCAs, where high capacity can be achieved using multiple parallel low-order OAM-modes instead of all high-order OAM-modes, to increase the capacity of transmitter-receiver aligned RowComms. In particular, we study two cases: 1). Concentric UCAs based RowComms without co-mode interference; 2). Concentric UCAs based RowComms with co-mode interference. We propose a co-mode-interference-free and a co-mode-interference-contained mode-decomposition schemes to recover transmit signals for these two cases, respectively. Also, we develop the optimal power allocation schemes to maximize the capacity for these two cases. Numerical simulations are presented to validate and evaluate that our developed concentric UCAs based low-order RowComms can significantly increase the capacity as compared with that of singular UCA based RowComms.

The plane wave based wireless communications have becoming more and more matured, along with the well utilization of the traditional resources such as time and frequency. To further increase the capacity for rapidly increasing capacity demand of wireless communications, it is potential to use the twist wave, which has the orbital angular momentum (OAM). In this paper, we discuss the OAM based wireless communications in the aspect of orthogonality, degree of freedom (DoF), and capacity, where both the transmitter and the receiver use uniform circular array (UCA) antennas. In particular, we compare OAM based wireless communications with multiple-input-multiple-output (MIMO) based wireless communications in terms of DoF and capacity. Numerical results are presented to validate and evaluate that the DoF of OAM based wireless communications is greater than or equal to that of correlated MIMO based wireless communications when the transmitter and the receiver antennas are aligned well. The OAM based wireless communications can achieve larger capacity than the correlated MIMO in high signal-to-noise ratio (SNR) region under line-of-sight scenario.

By enabling very high bandwidth for radio communications, the millimeter-wave (mmWave), which can easily be integrated with massive-multiple-input-multiple-output (massive-MIMO) due to small antenna size, has been attracting growing attention as a candidate for the fifth-generation (5G) and 5G-beyond wireless communications networks. On the other hand, the communication over the orthogonal states/modes of orbital angular momentum (OAM) is a subset of the solutions offered by massive-MIMO communications. Traditional massive-MIMO based mmWave communications did not concern the potential spectrum-efficiency-gain (SE-gain) offered by orthogonal states of OAM. However, the highly expecting maximum SE-gain for OAM and massive-MIMO communications is the product of SE-gains offered by OAM and multiplexing-MIMO. In this paper, we propose the OAM-embedded-MIMO (OEM) communication framework to obtain the multiplicative SE-gain for joint OAM and massive-MIMO based mmWave wireless communications. We design the parabolic antenna for each uniform circular array antenna to converge OAM signals. Then, we develop the mode-decomposition and multiplexing-detection scheme to obtain the transmit signal on each OAM-mode of each transmit antenna. Also, we develop the OEM-water-filling power allocation policy to achieve the maximum multiplicative SE-gain for OEM communications. The extensive simulations obtained validate and evaluate our developed parabolic antenna based converging method, mode-decomposition and multiplexing-detection scheme, and OEM-water-filling policy, showing that our proposed OEM mmWave communications can significantly increase the spectrum-efficiency as compared with traditional massive-MIMO based mmWave communications.

The development of orbital angular momentum (OAM)-based radio vortex transmission presents a promising opportunity for increasing the capacity of wireless communication in correlated channels due to its inherent orthogonality among different OAM modes. One of the most popular schemes for high-efficient OAM transmission is the digital baseband associated with uniform circular array (UCA) based transceiver. However, the periodicity of complex-exponential feed makes the maximum number of orthogonal signals carried by multiple OAM modes generally restricted to the array-element number of UCA antenna, which poses an open question of how to employ more OAM modes given a fixed number of array elements. Furthermore, signals modulated with high-order OAM modes are difficult to be captured by the receiver due to their serious divergence as propagating in free space, thus severely limiting the capacity of radio vortex communications. To overcome the above challenges, in this paper based on the partly element-overlapped fractal geometry layout and effectively using low-order OAM modes, we propose the quasi-fractal UCA (QF-UCA) antenna based OAM multiplexing transmission. We perform the two-dimension OAM modulation (TOM) and demodulation (TOD) schemes with the orthogonal OAM mode number exceeding the array-element number, which is beyond the traditional concept of multiple antennas based wireless communications. Simulation results show that our proposed scheme can achieve more number of orthogonal multiplexing streams than the maximum number of orthogonal multiplexing corresponding to traditional multiple antenna systems.

This paper proposes a bidirectional multibeam hybrid transmitarray (HTA) employing a transmission polarization-rotating metasurface (TPRM). A novel configuration is introduced to facilitate bidirectional beam scanning by combining the transmitarray (TA) and folded-transmitarray (FTA). To accomplish the reconfiguration of both unidirectional and bidirectional radiation states in the +z, -z, and +/-z directions, a polarization switchable multi-feed array (MFA) is placed at the focal plane between the TA and FTA, radiating x-polarization, y-polarization, and 45-degree oblique polarization waves, respectively. Meanwhile, the proposed antenna can achieve multibeam radiation in the three aforementioned states by switching the polarization of the MFA. To demonstrate the operating principle, a prototype has been designed, simulated, and fabricated. The measured results agree well with the simulated results. The simulated and measured results indicate that the proposed design can generate reconfigurable multibeam in both forward and backward directions, either separately or simultaneously. In the unidirectional states, forward and backward beam scanning is achieved within an angular range of +/-30{\deg} and +/-22{\deg}, respectively, with peak gains of 23.6 dBi and 23.1 dBi. A simultaneous forward and backward beam scanning of +/-40{\deg} and +/-22{\deg} is achieved in the hybrid radiation state, with peak gains of 19.4 dBi and 19.3 dBi, respectively. The proposed antenna array design offers several advantages, including bidirectional low-loss beam scanning, a simple structure, low power consumption, and a low profile., Comment: 12 pages, 26 figures, published to TAP

For unforeseen natural disasters, such as earthquakes, hurricanes, and floods, etc., the traditional communication infrastructure is unavailable or seriously disrupted along with persistent secondary disasters. Under such circumstances, it is highly demanded to deploy emergency wireless communication (EWC) networks to restore connectivity in accident/incident areas. The emerging fifth-generation (5G)/beyond-5G (B5G) wireless communication system, like unmanned aerial vehicle (UAV) assisted networks and intelligent reflecting surface (IRS) based communication systems, are expected to be designed or re-farmed for supporting temporary high quality communications in post-disaster areas. However, the channel characteristics of post-disaster areas quickly change as the secondary disaster resulted topographical changes, imposing new but critical challenges for EWC networks. In this paper, we propose a novel heterogeneous $\mathcal{F}$ composite fading channel model for EWC networks which accurately models and characterizes the composite fading channel with reflectors, path-loss exponent, fading, and shadowing parameters in 5G-UAV based EWC networks. Based on the model, we develop the optimal power allocation scheme with the simple closed-form expression and the numerical results based optimal joint bandwidth-power allocation scheme. We derive the corresponding capacities and compare the energy efficiency between IRS and traditional relay based 5G-UAVs. Numerical results show that the new heterogeneous Fisher-Snedecor $\mathcal{F}$ composite fading channel adapted resource allocation schemes can achieve higher capacity and energy efficiency than those of traditional channel model adapted resource allocation schemes, thus providing better communications service for post-disaster areas.

For unforeseen emergencies, such as natural disasters and pandemic events, it is highly demanded to cope with the explosive growth of mobile data traffic in extremely critical environments. An Unmanned aerial vehicle (UAV) fleet is an effective way to facilitate the Emergency wireless COmmunication NETwork (EcoNet). In this article, a MUlti-tier Heterogeneous UAV Network (MuHun), which is with different UAV fleets in different altitudes, is proposed to flexibly serve various emergencies. We refresh the key performance indicators of full coverage, network capacity, low latency, and energy efficiency in harsh environments. Then, we present the special challenges regarding shadowing-dominated complex channel model, energy supply limited short-endurance, various communication mechanisms coexistence, and communication island for underground users in UAV-based EcoNet, followed by the MuHun-based EcoNet architecture and its advantages. Furthermore, some potential solutions such as the new hybrid-channel adapted resource allocation, reconfigurable intelligent surface assisted UAV communications, competitive heterogenous-networks, and magnetic induction based air-to-ground/underground communications are discussed to effectively achieve full coverage, high capacity, high energy efficiency, and diverse qualities of services for EcoNets in harsh environments.

In this paper, we propose a virtual full-duplex (VFD) technique with zero-interval modulation and sampling (ZIMS), where two half-duplex (HD) transceivers can simultaneously transmit signals and each transceiver can effectively receive the desired information. In ZIMS-VFD, the transceiver inserts a zero-interval for each symbol in the transmit signal and provides self-interference (SI)-free intervals for itself. Meanwhile, it samples the receive signal in the provided SI-free intervals and restores the desired symbols. Based on orthogonal frequency division multiplexing (OFDM), we formulate the system model and show the transmit signal structure. Then, we give the transceiver design for single input single output (SISO) ZIMS-VFD and extend it to multiple input multiple output (MIMO) communications. Numerical results verify our theoretical analyses and show that ZIMS-VFD can effectively increase the capacity and approach the FD without SI.

Frequency hopping (FH) has been widely used as a powerful technique for antijamming in wireless communications. However, as the wireless spectrum is becoming more and more crowded, it is very difficult to achieve efficient antijamming results with FH-based schemes. Orbital angular momentum (OAM), which provides the new angular/mode dimension for wireless communications, offers an intriguing way for antijamming. In this paper, we propose to use the orthogonality of OAM-modes for antijamming in wireless communications. In particular, we propose the mode hopping (MH) scheme for antijamming within the narrow frequency band. We derive the closed-form expression of bit error rate (BER) for multiple users scenario with our developed MH scheme. Our developed MH scheme can achieve the same antijamming results within the narrow frequency band as compared with the conventional wideband FH scheme. Furthermore, we propose mode-frequency hopping (MFH) scheme, which jointly uses our developed MH scheme and the conventional FH scheme to further decrease the BER for wireless communication. Numerical results are presented to show that the BER of our developed MH scheme within the narrow frequency band is the same with that of the conventional wideband FH scheme. Moreover, the BER of our developed MFH schemes is much smaller than that of the conventional FH schemes for wireless communications., Comment: 15 pages, accepted by IEEE Transactions on Vehicular Technology (Volume: 67, Issue: 8, August 2018)

The emerging orbital angular momentum (OAM) based wireless communication is expected to be a high spectrum-efficiency communication paradigm to solve the growing transmission data rate and limited bandwidth problem. Academic researchers mainly concentrate on the OAM-based line-of-sight (LoS) communications. However, there exist some surroundings around the transceiver in most practical wireless communication scenarios, thus forming multipath transmission. In this paper, a hybrid orthogonal division multiplexing (HODM) scheme by using OAM multiplexing and orthogonal frequency division multiplexing (OFDM) in conjunction is proposed to achieve high-capacity wireless communications in sparse multipath environments, where the scatterers are sparse. We first build the OAM-based wireless channel in a LoS path and several reflection paths combined sparse multipath environments. We concentrate on less than or equal to three-time reflection paths because of the severe energy attenuation. The phase difference among the channel amplitude gains of the LoS and reflection paths, which is caused by the reflection paths, makes it difficult to decompose the OAM signals. We propose the phase difference compensation to handle this problem and then calculate the corresponding capacity in radio vortex wireless communications. Numerical results illustrate that the capacity of wireless communications by using our proposed HODM scheme can be drastically increased in sparse multipath environments., Comment: 15 pages, 12 figures, accepted by IEEE Transactions on Vehicular Technology ( Volume: 69, Issue: 4, April 2020). arXiv admin note: substantial text overlap with arXiv:1902.07542

Due to the crowded spectrum, frequency hopping (FH) techniques are now very difficult to achieve efficient antijamming and increase spectrum efficiency (SE) for wireless communications. The emerging orbital angular momentum (OAM), which is a property describing the helical phase fronts of electromagnetic waves, offers the potential to improve reliability and increase SE in wireless communications. To achieve efficient anti-jamming and increase SE of wireless communications with slight computational complexity cost, in this paper we propose an index-modulation embedded mode-hopping (IM-MH) scheme, which simultaneously activates several OAM-modes for hopping along with additional index information and signal information transmission. We analyze the average bit error rates (ABERs) for our proposed IM-MH scheme with perfect channel state information (CSI) and imperfect CSI, respectively. We also propose the index-modulation embedded double-serial MH (IMDSMH) scheme, which randomly activates one OAM-mode as the serial second hop to transmit the hopping signals in the IM-MH scheme, to further decrease the ABER of wireless communications. Extensive numerical results demonstrate that our proposed schemes within a narrowband can achieve the low ABER and significantly increase the SE. Also, the ABERs of our proposed IM-MH and IM-DSMH schemes are around 25% and 10%, respectively, compared with that of the mode hopping scheme., Comment: 13 pages, 12 figures, accepted by IEEE SYSTEMS JOURNAL

Orbital angular momentum (OAM) based mode hopping (MH) scheme is expected to be a potential anti-jamming technology in radio vortex wireless communications. However, it only uses one OAM-mode for hopping, thus resulting in low spectrum efficiency (SE). Index modulation offers a trade-off balance between the SE and performance reliability. In this paper, we propose an MH with OAM-based index modulation scheme, where several OAM-modes are activated for hopping, to achieve high SE at a given bit error rate in radio vortex wireless communications. Based on the proposed scheme, we derive the upper bound and lower bound of achievable SEs. Furthermore, in order to take advantage of index information, we derive the optimal hopped OAM-modes to achieve the maximum SE. Numerical results show that our proposed MH with index modulation scheme can achieve high SE while satisfying a certain reliability of radio vortex wireless communications., Comment: 7 pages, 5 figures, accepted by 2019 IEEE Global Communications Conference (GLOBECOM)

Nowadays, Large Language Models (LLMs) have been trained using extended context lengths to foster more creative applications. However, long context training poses great challenges considering the constraint of GPU memory. It not only leads to substantial activation memory consumption during training, but also incurs considerable memory fragmentation. To facilitate long context training, existing frameworks have adopted strategies such as recomputation and various forms of parallelisms. Nevertheless, these techniques rely on redundant computation or extensive communication, resulting in low Model FLOPS Utilization (MFU). In this paper, we propose MEMO, a novel LLM training framework designed for fine-grained activation memory management. Given the quadratic scaling of computation and linear scaling of memory with sequence lengths when using FlashAttention, we offload memory-consuming activations to CPU memory after each layer's forward pass and fetch them during the backward pass. To maximize the swapping of activations without hindering computation, and to avoid exhausting limited CPU memory, we implement a token-wise activation recomputation and swapping mechanism. Furthermore, we tackle the memory fragmentation issue by employing a bi-level Mixed Integer Programming (MIP) approach, optimizing the reuse of memory across transformer layers. Empirical results demonstrate that MEMO achieves an average of 2.42x and 2.26x MFU compared to Megatron-LM and DeepSpeed, respectively. This improvement is attributed to MEMO's ability to minimize memory fragmentation, reduce recomputation and intensive communication, and circumvent the delays associated with the memory reorganization process due to fragmentation. By leveraging fine-grained activation memory management, MEMO facilitates efficient training of 7B LLM with 1 million sequence length on just 8 A800 GPUs, achieving an MFU of 52.30%.

As a kind of wavefront with helical phase, orbital angular momentum (OAM) shows the great potential to enhance the security results of wireless communications due to its unique orthogonality and central hollow electromagnetic wave structure. Therefore, in this paper we propose the reconfigurable-intelligent-surface (RIS) assisted OAM scheme, where RIS is deployed to weaken the information acquisition at eavesdroppers by adjusting the OAM beams pointed to the eavesdropper and artificial noise (AN) is applied to interfere with the eavesdropper, thus significantly increasing the secrecy rates of short-range secure communications. Aiming at obtaining the maximum secrecy rate, we develop the Riemannian manifold conjugate gradient (RMCG) based alternative optimization (AO) algorithm to assign much power to low-order OAM-modes and optimize the OAM beams direction with the programmable RIS, thus respectively enhancing and weakening the received signal strength at the legitimate receiver and the eavesdropper. Numerical results show that our proposed scheme outperforms the existing works in terms of the secrecy rate and the eavesdropper's bit error rate., Comment: arXiv admin note: text overlap with arXiv:2406.05799

As the field of Large Language Models (LLMs) continues to evolve, the context length in inference is steadily growing. Key-Value Cache (KVCache), a crucial component in LLM inference, has now become the primary memory bottleneck due to limited GPU memory. Current methods selectively determine suitable keys and values for self-attention computation in LLMs to address the issue. However, they either fall short in maintaining model quality or result in high serving latency. Drawing inspiration from advanced embedding retrieval techniques used in the database community, we consider the storage and searching of KVCache as a typical embedding retrieval problem. We propose PQCache, which employs Product Quantization (PQ) to manage KVCache, maintaining model quality while ensuring low serving latency. During the prefilling phase, we apply PQ to tokens' keys for each LLM layer and head. During the autoregressive decoding phase, for each newly generated token, we first identify important tokens through Maximum Inner-Product Search (MIPS) using PQ codes and centroids, then fetch the corresponding key-value pairs for self-attention computation. Through meticulous design of overlapping and caching, we minimize any additional computation and communication overhead during both phases. Extensive experiments show that PQCache achieves both effectiveness and efficiency. It maintains model quality even when only 1/5 of the tokens are involved in attention, while attaining acceptable system latency.

Simultaneous wireless information and power transfer (SWIPT), which achieves both wireless energy transfer (WET) and information transfer, is an attractive technique for future Internet of Everything (IoE) in the sixth-generation (6G) mobile communications. With SWIPT, battery-less IoE devices can be powered while communicating with other devices. Line-of-sight (LOS) RF transmission and near-field inductive coupling based transmission are typical SWIPT scenarios, which are both LOS channels and without enough degree of freedom for high spectrum efficiency as well as high energy efficiency. Due to the orthogonal wavefronts, orbital angular momentum (OAM) can facilitate the SWIPT in LOS channels. In this article, we introduce the OAM-based SWIPT as well as discuss some basic advantages and challenges for it. After introducing the OAM-based SWIPT for IoE, we first propose an OAM-based SWIPT system model with the OAM-modes assisted dynamic power splitting (DPS). Then, four basic advantages regarding the OAM-based SWIPT are reviewed with some numerical analyses for further demonstrating the advantages. Next, four challenges regarding integrating OAM into SWIPT and possible solutions are discussed. OAM technology provides multiple orthogonal streams to increase both spectrum and energy efficiencies for SWIPT, thus creating many opportunities for future WET and SWIPT researches., Comment: 7 pages, 6 figures

In current deep learning tasks, Adam style optimizers such as Adam, Adagrad, RMSProp, Adafactor, and Lion have been widely used as alternatives to SGD style optimizers. These optimizers typically update model parameters using the sign of gradients, resulting in more stable convergence curves. The learning rate and the batch size are the most critical hyperparameters for optimizers, which require careful tuning to enable effective convergence. Previous research has shown that the optimal learning rate increases linearly or follows similar rules with batch size for SGD style optimizers. However, this conclusion is not applicable to Adam style optimizers. In this paper, we elucidate the connection between optimal learning rates and batch sizes for Adam style optimizers through both theoretical analysis and extensive experiments. First, we raise the scaling law between batch sizes and optimal learning rates in the sign of gradient case, in which we prove that the optimal learning rate first rises and then falls as the batch size increases. Moreover, the peak value of the surge will gradually move toward the larger batch size as training progresses. Second, we conducted experiments on various CV and NLP tasks and verified the correctness of the scaling law.

Advancements in model algorithms, the growth of foundational models, and access to high-quality datasets have propelled the evolution of Artificial Intelligence Generated Content (AIGC). Despite its notable successes, AIGC still faces hurdles such as updating knowledge, handling long-tail data, mitigating data leakage, and managing high training and inference costs. Retrieval-Augmented Generation (RAG) has recently emerged as a paradigm to address such challenges. In particular, RAG introduces the information retrieval process, which enhances the generation process by retrieving relevant objects from available data stores, leading to higher accuracy and better robustness. In this paper, we comprehensively review existing efforts that integrate RAG technique into AIGC scenarios. We first classify RAG foundations according to how the retriever augments the generator, distilling the fundamental abstractions of the augmentation methodologies for various retrievers and generators. This unified perspective encompasses all RAG scenarios, illuminating advancements and pivotal technologies that help with potential future progress. We also summarize additional enhancements methods for RAG, facilitating effective engineering and implementation of RAG systems. Then from another view, we survey on practical applications of RAG across different modalities and tasks, offering valuable references for researchers and practitioners. Furthermore, we introduce the benchmarks for RAG, discuss the limitations of current RAG systems, and suggest potential directions for future research. Github: https://github.com/PKU-DAIR/RAG-Survey., Comment: Citing 353 papers, 22 pages, 1 table, 12 figures. Project: https://github.com/PKU-DAIR/RAG-Survey

Recently, the growing memory demands of embedding tables in Deep Learning Recommendation Models (DLRMs) pose great challenges for model training and deployment. Existing embedding compression solutions cannot simultaneously meet three key design requirements: memory efficiency, low latency, and adaptability to dynamic data distribution. This paper presents CAFE, a Compact, Adaptive, and Fast Embedding compression framework that addresses the above requirements. The design philosophy of CAFE is to dynamically allocate more memory resources to important features (called hot features), and allocate less memory to unimportant ones. In CAFE, we propose a fast and lightweight sketch data structure, named HotSketch, to capture feature importance and report hot features in real time. For each reported hot feature, we assign it a unique embedding. For the non-hot features, we allow multiple features to share one embedding by using hash embedding technique. Guided by our design philosophy, we further propose a multi-level hash embedding framework to optimize the embedding tables of non-hot features. We theoretically analyze the accuracy of HotSketch, and analyze the model convergence against deviation. Extensive experiments show that CAFE significantly outperforms existing embedding compression methods, yielding 3.92% and 3.68% superior testing AUC on Criteo Kaggle dataset and CriteoTB dataset at a compression ratio of 10000x. The source codes of CAFE are available at GitHub.

Learnable embedding vector is one of the most important applications in machine learning, and is widely used in various database-related domains. However, the high dimensionality of sparse data in recommendation tasks and the huge volume of corpus in retrieval-related tasks lead to a large memory consumption of the embedding table, which poses a great challenge to the training and deployment of models. Recent research has proposed various methods to compress the embeddings at the cost of a slight decrease in model quality or the introduction of other overheads. Nevertheless, the relative performance of these methods remains unclear. Existing experimental comparisons only cover a subset of these methods and focus on limited metrics. In this paper, we perform a comprehensive comparative analysis and experimental evaluation of embedding compression. We introduce a new taxonomy that categorizes these techniques based on their characteristics and methodologies, and further develop a modular benchmarking framework that integrates 14 representative methods. Under a uniform test environment, our benchmark fairly evaluates each approach, presents their strengths and weaknesses under different memory budgets, and recommends the best method based on the use case. In addition to providing useful guidelines, our study also uncovers the limitations of current methods and suggests potential directions for future research.

Embedding-based retrieval methods construct vector indices to search for document representations that are most similar to the query representations. They are widely used in document retrieval due to low latency and decent recall performance. Recent research indicates that deep retrieval solutions offer better model quality, but are hindered by unacceptable serving latency and the inability to support document updates. In this paper, we aim to enhance the vector index with end-to-end deep generative models, leveraging the differentiable advantages of deep retrieval models while maintaining desirable serving efficiency. We propose Model-enhanced Vector Index (MEVI), a differentiable model-enhanced index empowered by a twin-tower representation model. MEVI leverages a Residual Quantization (RQ) codebook to bridge the sequence-to-sequence deep retrieval and embedding-based models. To substantially reduce the inference time, instead of decoding the unique document ids in long sequential steps, we first generate some semantic virtual cluster ids of candidate documents in a small number of steps, and then leverage the well-adapted embedding vectors to further perform a fine-grained search for the relevant documents in the candidate virtual clusters. We empirically show that our model achieves better performance on the commonly used academic benchmarks MSMARCO Passage and Natural Questions, with comparable serving latency to dense retrieval solutions.

Multi-Teacher knowledge distillation provides students with additional supervision from multiple pre-trained teachers with diverse information sources. Most existing methods explore different weighting strategies to obtain a powerful ensemble teacher, while ignoring the student with poor learning ability may not benefit from such specialized integrated knowledge. To address this problem, we propose Adaptive Multi-teacher Knowledge Distillation with Meta-Learning (MMKD) to supervise student with appropriate knowledge from a tailored ensemble teacher. With the help of a meta-weight network, the diverse yet compatible teacher knowledge in the output layer and intermediate layers is jointly leveraged to enhance the student performance. Extensive experiments on multiple benchmark datasets validate the effectiveness and flexibility of our methods. Code is available: https://github.com/Rorozhl/MMKD., Comment: ICME 2023

Transformer models have achieved state-of-the-art performance on various domains of applications and gradually becomes the foundations of the advanced large deep learning (DL) models. However, how to train these models over multiple GPUs efficiently is still challenging due to a large number of parallelism choices. Existing DL systems either rely on manual efforts to make distributed training plans or apply parallelism combinations within a very limited search space. In this approach, we propose Galvatron, a new system framework that incorporates multiple popular parallelism dimensions and automatically finds the most efficient hybrid parallelism strategy. To better explore such a rarely huge search space, we 1) involve a decision tree to make decomposition and pruning based on some reasonable intuitions, and then 2) design a dynamic programming search algorithm to generate the optimal plan. Evaluations on four representative Transformer workloads show that Galvatron could perform automatically distributed training with different GPU memory budgets. Among all evluated scenarios, Galvatron always achieves superior system throughput compared to previous work with limited parallelism.

Knowledge distillation aims to compress a powerful yet cumbersome teacher model into a lightweight student model without much sacrifice of performance. For this purpose, various approaches have been proposed over the past few years, generally with elaborately designed knowledge representations, which in turn increase the difficulty of model development and interpretation. In contrast, we empirically show that a simple knowledge distillation technique is enough to significantly narrow down the teacher-student performance gap. We directly reuse the discriminative classifier from the pre-trained teacher model for student inference and train a student encoder through feature alignment with a single $\ell_2$ loss. In this way, the student model is able to achieve exactly the same performance as the teacher model provided that their extracted features are perfectly aligned. An additional projector is developed to help the student encoder match with the teacher classifier, which renders our technique applicable to various teacher and student architectures. Extensive experiments demonstrate that our technique achieves state-of-the-art results at the modest cost of compression ratio due to the added projector., Comment: Accepted to CVPR-2022

Due to the limited computing resources of swarm of drones, it is difficult to handle computation-intensive tasks locally, hence the cloud based computation offloading is widely adopted. However, for the business which requires low latency and high reliability, the cloud-based solution is not suitable, because of the slow response time caused by long distance data transmission. Therefore, to solve the problem mentioned above, in this paper, we introduce fog computing into swarm of drones (FCSD). Focusing on the latency and reliability sensitive business scenarios, the latency and reliability is constructed as the constraints of the optimization problem. And in order to enhance the practicality of the FCSD system, we formulate the energy consumption of FCSD as the optimization target function, to decrease the energy consumption as far as possible, under the premise of satisfying the latency and reliability requirements of the task. Furthermore, a heuristic algorithm based on genetic algorithm is designed to perform optimal task allocation in FCSD system. The simulation results validate that the proposed fog based computation offloading with the heuristic algorithm can complete the computing task effectively with the minimal energy consumption under the requirements of latency and reliability.