The commonly used caching policies, such as LRU or LFU, exhibit optimal performance only for specific traffic patterns. Even advanced Machine Learning-based methods, which detect patterns in historical request data, struggle when future requests deviate from past trends. Recently, a new class of policies has emerged that makes no assumptions about the request arrival process. These algorithms solve an online optimization problem, enabling continuous adaptation to the context. They offer theoretical guarantees on the regret metric, which is the gap between the gain of the online policy and the gain of the optimal static cache allocation in hindsight. Nevertheless, the high computational complexity of these solutions hinders their practical adoption. In this study, we introduce a groundbreaking gradient-based online caching policy, the first to achieve logarithmic computational complexity relative to catalog size along with regret guarantees. This means our algorithm can efficiently handle large-scale data while minimizing the performance gap between real-time decisions and optimal hindsight choices. As requests arrive, our policy dynamically adjusts the probabilities of including items in the cache, which drive cache update decisions. Our algorithm's streamlined complexity is a key advantage, enabling its application to real-world traces featuring millions of requests and items. This is a significant achievement, as traces of this scale have been out of reach for existing policies with regret guarantees. To the best of our knowledge, our experimental results show for the first time that the regret guarantees of gradient-based caching policies bring significant benefits in scenarios of practical interest.
相關內容
Non-autoregressive (NAR) automatic speech recognition (ASR) models predict tokens independently and simultaneously, bringing high inference speed. However, there is still a gap in the accuracy of the NAR models compared to the autoregressive (AR) models. To further narrow the gap between the NAR and AR models, we propose a single-step NAR ASR architecture with high accuracy and inference speed, called EfficientASR. It uses an Index Mapping Vector (IMV) based alignment generator to generate alignments during training, and an alignment predictor to learn the alignments for inference. It can be trained end-to-end (E2E) with cross-entropy loss combined with alignment loss. The proposed EfficientASR achieves competitive results on the AISHELL-1 and AISHELL-2 benchmarks compared to the state-of-the-art (SOTA) models. Specifically, it achieves character error rates (CER) of 4.26%/4.62% on the AISHELL-1 dev/test dataset, which outperforms the SOTA AR Conformer with about 30x inference speedup.
Dropped Head Syndrome (DHS) causes a passively correctable neck deformation. Currently, there is no wearable orthopedic neck brace to fulfill the needs of persons suffering from DHS. Related works have made progress in this area by creating mobile neck braces that provide head support to mitigate deformation while permitting neck mobility, which enhances user-perceived comfort and quality of life. Specifically, passive designs show great potential for fully functional devices in the short term due to their inherent simplicity and compactness, although achieving suitable support presents some challenges. This work introduces a novel compliant mechanism that provides non-restrictive adjustable support for the neck's anterior and posterior flexion movements while enabling its unconstrained free rotation. The results from the experiments on non-affected persons suggest that the device provides the proposed adjustable support that unloads the muscle groups involved in supporting the head without overloading the antagonist muscle groups. Simultaneously, it was verified that the free rotation is achieved regardless of the stiffness configuration of the device.
Efficiency in instruction fetching is critical to performance, and this requires the primary structures -- L1 instruction caches (L1i), branch target buffers (BTB) and instruction TLBs (iTLB) -- to have the requisite information when needed. This paper proposes a high-level program sequencing mechanism and a coupled technique for block movement, instruction presending, where instruction cache blocks, BTB entries, and iTLB entries are autonomously moved (or sent) from the secondary to the primary structures in a "just in time" fashion so that they are available when needed. Empirical results are presented to demonstrate the efficacy of the high-level sequencing mechanism and block movement. Presending is especially effective for benchmarks with a high base MPKI, where the movement of instruction blocks (and BTB/iTLB entries) from secondary to primary structures is frequent. Presending reduces the number of misses in primary structures by an order of magnitude as compared to state-of-the-art instruction prefetching schemes, in many cases, while allowing the processor to operate with small-sized primary BTBs. This reduction in misses results in performance improvements in cases where front-end efficiency is important.
Speech Self-Supervised Learning (SSL) has demonstrated considerable efficacy in various downstream tasks. Nevertheless, prevailing self-supervised models often overlook the incorporation of emotion-related prior information, thereby neglecting the potential enhancement of emotion task comprehension through emotion prior knowledge in speech. In this paper, we propose an emotion-aware speech representation learning with intensity knowledge. Specifically, we extract frame-level emotion intensities using an established speech-emotion understanding model. Subsequently, we propose a novel emotional masking strategy (EMS) to incorporate emotion intensities into the masking process. We selected two representative models based on Transformer and CNN, namely MockingJay and Non-autoregressive Predictive Coding (NPC), and conducted experiments on IEMOCAP dataset. Experiments have demonstrated that the representations derived from our proposed method outperform the original model in SER task.
Transformer models can face practical limitations due to their high computational requirements. At the same time, such models exhibit significant activation sparsity, which can be leveraged to reduce the inference cost by converting parts of the network into equivalent Mixture-of-Experts (MoE) layers. Despite the crucial role played by activation sparsity, its impact on this process remains unexplored. In particular, we show that the efficiency of the conversion can be significantly enhanced by a proper regularization of the activation sparsity of the base model. Moreover, motivated by the high variance of the number of activated neurons for different inputs, we introduce a more effective dynamic-k expert selection rule that adjusts the number of executed experts on a per-token basis. Finally, we extend this approach to multi-head attention projections, which results in additional savings compared to only converting the FFN blocks. The proposed method, Dense to Dynamic-$k$ Mixture-of-Experts (D2DMoE), outperforms existing approaches on common NLP and vision tasks, allowing us to save up to 60% of inference cost without significantly affecting model performance.
Despite recent improvements in End-to-End Automatic Speech Recognition (E2E ASR) systems, the performance can degrade due to vocal characteristic mismatches between training and testing data, particularly with limited target speaker adaptation data. We propose a novel speaker adaptation approach Speaker-Smoothed kNN that leverages k-Nearest Neighbors (kNN) retrieval techniques to improve model output by finding correctly pronounced tokens from its pre-built datastore during the decoding phase. Moreover, we utilize x-vector to dynamically adjust kNN interpolation parameters for data sparsity issue. This approach was validated using KeSpeech and MagicData corpora under in-domain and all-domain settings. Our method consistently performs comparably to fine-tuning without the associated performance degradation during speaker changes. Furthermore, in the all-domain setting, our method achieves state-of-the-art results, reducing the CER in both single speaker and multi-speaker test scenarios.
Adversarial training has achieved substantial performance in defending image retrieval against adversarial examples. However, existing studies in deep metric learning (DML) still suffer from two major limitations: weak adversary and model collapse. In this paper, we address these two limitations by proposing Collapse-Aware TRIplet DEcoupling (CA-TRIDE). Specifically, TRIDE yields a stronger adversary by spatially decoupling the perturbation targets into the anchor and the other candidates. Furthermore, CA prevents the consequential model collapse, based on a novel metric, collapseness, which is incorporated into the optimization of perturbation. We also identify two drawbacks of the existing robustness metric in image retrieval and propose a new metric for a more reasonable robustness evaluation. Extensive experiments on three datasets demonstrate that CA-TRIDE outperforms existing defense methods in both conventional and new metrics. Codes are available at //github.com/michaeltian108/CA-TRIDE.
Is it possible accurately classify political relations within evolving event ontologies without extensive annotations? This study investigates zero-shot learning methods that use expert knowledge from existing annotation codebook, and evaluates the performance of advanced ChatGPT (GPT-3.5/4) and a natural language inference (NLI)-based model called ZSP. ChatGPT uses codebook's labeled summaries as prompts, whereas ZSP breaks down the classification task into context, event mode, and class disambiguation to refine task-specific hypotheses. This decomposition enhances interpretability, efficiency, and adaptability to schema changes. The experiments reveal ChatGPT's strengths and limitations, and crucially show ZSP's outperformance of dictionary-based methods and its competitive edge over some supervised models. These findings affirm the value of ZSP for validating event records and advancing ontology development. Our study underscores the efficacy of leveraging transfer learning and existing domain expertise to enhance research efficiency and scalability.
Automatic KB completion for commonsense knowledge graphs (e.g., ATOMIC and ConceptNet) poses unique challenges compared to the much studied conventional knowledge bases (e.g., Freebase). Commonsense knowledge graphs use free-form text to represent nodes, resulting in orders of magnitude more nodes compared to conventional KBs (18x more nodes in ATOMIC compared to Freebase (FB15K-237)). Importantly, this implies significantly sparser graph structures - a major challenge for existing KB completion methods that assume densely connected graphs over a relatively smaller set of nodes. In this paper, we present novel KB completion models that can address these challenges by exploiting the structural and semantic context of nodes. Specifically, we investigate two key ideas: (1) learning from local graph structure, using graph convolutional networks and automatic graph densification and (2) transfer learning from pre-trained language models to knowledge graphs for enhanced contextual representation of knowledge. We describe our method to incorporate information from both these sources in a joint model and provide the first empirical results for KB completion on ATOMIC and evaluation with ranking metrics on ConceptNet. Our results demonstrate the effectiveness of language model representations in boosting link prediction performance and the advantages of learning from local graph structure (+1.5 points in MRR for ConceptNet) when training on subgraphs for computational efficiency. Further analysis on model predictions shines light on the types of commonsense knowledge that language models capture well.
Dense video captioning aims to generate text descriptions for all events in an untrimmed video. This involves both detecting and describing events. Therefore, all previous methods on dense video captioning tackle this problem by building two models, i.e. an event proposal and a captioning model, for these two sub-problems. The models are either trained separately or in alternation. This prevents direct influence of the language description to the event proposal, which is important for generating accurate descriptions. To address this problem, we propose an end-to-end transformer model for dense video captioning. The encoder encodes the video into appropriate representations. The proposal decoder decodes from the encoding with different anchors to form video event proposals. The captioning decoder employs a masking network to restrict its attention to the proposal event over the encoding feature. This masking network converts the event proposal to a differentiable mask, which ensures the consistency between the proposal and captioning during training. In addition, our model employs a self-attention mechanism, which enables the use of efficient non-recurrent structure during encoding and leads to performance improvements. We demonstrate the effectiveness of this end-to-end model on ActivityNet Captions and YouCookII datasets, where we achieved 10.12 and 6.58 METEOR score, respectively.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191