The deep learning community has witnessed an exponentially growing interest in self-supervised learning (SSL). However, it still remains unexplored how to build a framework for learning useful representations of raw music waveforms in a self-supervised manner. In this work, we design Music2Vec, a framework exploring different SSL algorithmic components and tricks for music audio recordings. Our model achieves comparable results to the state-of-the-art (SOTA) music SSL model Jukebox, despite being significantly smaller with less than 2% of parameters of the latter. The model will be released on Huggingface(Please refer to: https://huggingface.co/m-a-p/music2vec-v1)

自我监督的学习（SSL）从大量未标记的数据中学习知识，然后将知识转移到有限数量的标记数据的特定问题上。SSL在各个领域都取得了有希望的结果。这项工作解决了细分级通用音频SSL的问题，并提出了一个新的基于变压器的教师学生SSL模型，名为ATST。在最近出现的教师基线方案上开发了变压器编码器，该方案在很大程度上提高了预训练的建模能力。此外，旨在充分利用变压器的能力的新策略旨在充分利用。已经进行了广泛的实验，并且提出的模型几乎所有下游任务都实现了新的最新结果。

音乐信息检索的音频表示通常通过以特定于任务的方式通过监督学习来学习。虽然有效地产生最先进的结果，但该方案对于模型可以具有并且需要广泛的注释数据集的应用范围缺乏灵活性。在这项工作中，我们构成了是否可以利用弱对齐文本作为唯一用于学习通用音频音频表示的监督信号的问题。为了解决这个问题，我们设计了通过一组代理任务优化的音乐和语言预训练（Mulap）的多模式架构。弱监管以嘈杂的自然语言描述形式传达轨道的整体音乐纪念。在预训练之后，我们将模型的音频骨干转换为一组音乐音频分类和回归任务。我们通过比较通过不同培训策略产生的相同音频骨干声音产生的音频表示的性能并表明我们的预训练方法始终如一地实现所有任务和数据集所考虑的可比分数，因此证明了我们的方法。我们的实验还证实，Mulap有效利用音频标题对，以学习与文献中的音频和跨模型自我监督方法具有竞争力的表示。

In recent years, the development of accurate deep keyword spotting (KWS) models has resulted in KWS technology being embedded in a number of technologies such as voice assistants. Many of these models rely on large amounts of labelled data to achieve good performance. As a result, their use is restricted to applications for which a large labelled speech data set can be obtained. Self-supervised learning seeks to mitigate the need for large labelled data sets by leveraging unlabelled data, which is easier to obtain in large amounts. However, most self-supervised methods have only been investigated for very large models, whereas KWS models are desired to be small. In this paper, we investigate the use of self-supervised pretraining for the smaller KWS models in a label-deficient scenario. We pretrain the Keyword Transformer model using the self-supervised framework Data2Vec and carry out experiments on a label-deficient setup of the Google Speech Commands data set. It is found that the pretrained models greatly outperform the models without pretraining, showing that Data2Vec pretraining can increase the performance of KWS models in label-deficient scenarios. The source code is made publicly available.

The massive growth of self-supervised learning (SSL) has been witnessed in language, vision, speech, and audio domains over the past few years. While discrete label prediction is widely adopted for other modalities, the state-of-the-art audio SSL models still employ reconstruction loss for pre-training. Compared with reconstruction loss, semantic-rich discrete label prediction encourages the SSL model to abstract the high-level audio semantics and discard the redundant details as in human perception. However, a semantic-rich acoustic tokenizer for general audio pre-training is usually not straightforward to obtain, due to the continuous property of audio and unavailable phoneme sequences like speech. To tackle this challenge, we propose BEATs, an iterative audio pre-training framework to learn Bidirectional Encoder representation from Audio Transformers, where an acoustic tokenizer and an audio SSL model are optimized by iterations. In the first iteration, we use random projection as the acoustic tokenizer to train an audio SSL model in a mask and label prediction manner. Then, we train an acoustic tokenizer for the next iteration by distilling the semantic knowledge from the pre-trained or fine-tuned audio SSL model. The iteration is repeated with the hope of mutual promotion of the acoustic tokenizer and audio SSL model. The experimental results demonstrate our acoustic tokenizers can generate discrete labels with rich audio semantics and our audio SSL models achieve state-of-the-art results across various audio classification benchmarks, even outperforming previous models that use more training data and model parameters significantly. Specifically, we set a new state-of-the-art mAP 50.6% on AudioSet-2M for audio-only models without using any external data, and 98.1% accuracy on ESC-50. The code and pre-trained models are available at https://aka.ms/beats.

这项工作介绍了Brillsson，这是一种基于二进制神经网络的新型表示模型，用于广泛的非语义语音任务。我们从一个大型且价值的琐事模型中使用知识蒸馏来训练该模型，其中仅用于训练Trillsson的数据集中只有一小部分。由此产生的Brillsson型号的尺寸仅为2MB，潜伏期小于8ms，使其适合在低资源设备（例如可穿戴设备）中部署。我们在八项基准任务（包括但不限于口语识别，情感识别，荒地状况诊断和关键字斑点）上评估布里尔森，并证明我们提出的拟议的超轻质和低延迟模型以及大型模型以及大型模型。

我们向2021年的情感和主题提出了可摩擦的提交。在这项工作中，我们打算解决问题：我们可以利用关于音乐情感认可的半监督学习技巧吗？有了，我们试验嘈杂的学生培训，在图像分类域中具有改进的模型性能。随着嘈杂的学生方法需要强大的教师模型，我们进一步深入研究（i）输入培训长度和（ii）互补音乐表示，以进一步提高教师模型的表现。对于（i），我们发现，在PR-AUC中，具有短输入长度的型号更好地执行，而在ROC-AUC中具有长输入长度的培训则更好地执行这些模型。对于（ii），我们发现，使用谐波俯仰类概况（HPCP）一致地提高标记性能，这表明谐波表示对于音乐情感标记是有用的。最后，我们发现嘈杂的学生方法只改善了长训练长度的情况的标记结果。此外，我们发现，使用不同训练长度培训的合奏表示可以显着提高标记结果，这表明探索在网络架构中的多个时间分辨率探索的可能方向。

Current self-supervised learning algorithms are often modality-specific and require large amounts of computational resources. To address these issues, we increase the training efficiency of data2vec, a learning objective that generalizes across several modalities. We do not encode masked tokens, use a fast convolutional decoder and amortize the effort to build teacher representations. data2vec 2.0 benefits from the rich contextualized target representations introduced in data2vec which enable a fast self-supervised learner. Experiments on ImageNet-1K image classification show that data2vec 2.0 matches the accuracy of Masked Autoencoders in 16.4x lower pre-training time, on Librispeech speech recognition it performs as well as wav2vec 2.0 in 10.6x less time, and on GLUE natural language understanding it matches a retrained RoBERTa model in half the time. Trading some speed for accuracy results in ImageNet-1K top-1 accuracy of 86.8\% with a ViT-L model trained for 150 epochs.

自我监督的语音表示学习在各种语音处理任务中显示出令人鼓舞的结果。但是，预先训练的模型，例如休伯特是存储密集型变压器，限制了其在低资源设置下的应用程序范围。为此，我们建议通过修剪结构化参数自动找到所需的体系结构Lighthubert，这是一个曾经是变压器的压缩框架。更确切地说，我们创建了一个基于变压器的超级网，该超网嵌套着数千个重量共享子网，并设计了一个两阶段的蒸馏策略，以利用休伯特的上下文化潜在表示。关于自动语音识别（ASR）和出色基准的实验表明，拟议的lighthubert可实现$ 10^9 $的架构，该体系结构涉及嵌入尺寸，注意力维度，头部编号，进率向前网络比率和网络深度。 Lighthubert优于ASR上的原始Hubert和Hubert大小的五个出色的任务，在大多数任务中，在大多数任务中都具有可比的性能，并减少了29％的参数，并获得了$ 3.5 \ times $ times $ compression $压缩比在三个超级任务中，例如自动扬声器验证，关键字发现和意图分类，略有准确的损失。代码和预培训模型可在https://github.com/mechanicalsea/lighthubert上找到。

Self-supervised approaches for speech representation learning are challenged by three unique problems: (1) there are multiple sound units in each input utterance, (2) there is no lexicon of input sound units during the pre-training phase, and (3) sound units have variable lengths with no explicit segmentation. To deal with these three problems, we propose the Hidden-Unit BERT (HuBERT) approach for self-supervised speech representation learning, which utilizes an offline clustering step to provide aligned target labels for a BERT-like prediction loss. A key ingredient of our approach is applying the prediction loss over the masked regions only, which forces the model to learn a combined acoustic and language model over the continuous inputs. HuBERT relies primarily on the consistency of the unsupervised clustering step rather than the intrinsic quality of the assigned cluster labels. Starting with a simple k-means teacher of 100 clusters, and using two iterations of clustering, the HuBERT model either matches or improves upon the state-ofthe-art wav2vec 2.0 performance on the Librispeech (960h) and Libri-light (60,000h) benchmarks with 10min, 1h, 10h, 100h, and 960h fine-tuning subsets. Using a 1B parameter model, HuBERT shows up to 19% and 13% relative WER reduction on the more challenging dev-other and test-other evaluation subsets. 1

大规模的语音自我监督学习（SSL）已经出现到语音处理的主要领域，但是，由于其巨大规模而引起的计算成本问题是对学术界的高障碍。此外，语音SSL模型的现有蒸馏技术通过减少层来压缩模型，从而在语言模式识别任务（例如音素识别（PR））中引起性能降解。在本文中，我们提出了Fithubert，它几乎在几乎所有模型组件中都使尺寸较薄，并且与先前的语音SSL蒸馏作品相比，层层更深。此外，我们采用缩短时间来加快推理时间，并提出一种基于提示的蒸馏方法，以减少性能降解。与休伯特相比，我们的方法将模型降低到23.8％，推理时间为35.9％。此外，我们在优越的基准上达到了12.1％的单词错误率和13.3％的音素错误率，这比先前的工作优越。

We present Masked Audio-Video Learners (MAViL) to train audio-visual representations. Our approach learns with three complementary forms of self-supervision: (1) reconstruction of masked audio and video input data, (2) intra- and inter-modal contrastive learning with masking, and (3) self-training by reconstructing joint audio-video contextualized features learned from the first two objectives. Pre-training with MAViL not only enables the model to perform well in audio-visual classification and retrieval tasks but also improves representations of each modality in isolation, without using information from the other modality for fine-tuning or inference. Empirically, MAViL sets a new state-of-the-art on AudioSet (53.1 mAP) and VGGSound (67.1% accuracy). For the first time, a self-supervised audio-visual model outperforms ones that use external supervision on these benchmarks. Code will be available soon.

We present RAVEn, a self-supervised multi-modal approach to jointly learn visual and auditory speech representations. Our pre-training objective involves encoding masked inputs, and then predicting contextualised targets generated by slowly-evolving momentum encoders. Driven by the inherent differences between video and audio, our design is asymmetric w.r.t. the two modalities' pretext tasks: Whereas the auditory stream predicts both the visual and auditory targets, the visual one predicts only the auditory targets. We observe strong results in low- and high-resource labelled data settings when fine-tuning the visual and auditory encoders resulting from a single pre-training stage, in which the encoders are jointly trained. Notably, RAVEn surpasses all self-supervised methods on visual speech recognition (VSR) on LRS3, and combining RAVEn with self-training using only 30 hours of labelled data even outperforms a recent semi-supervised method trained on 90,000 hours of non-public data. At the same time, we achieve state-of-the-art results in the LRS3 low-resource setting for auditory speech recognition (as well as for VSR). Our findings point to the viability of learning powerful speech representations entirely from raw video and audio, i.e., without relying on handcrafted features. Code and models will be made public.

我们利用Libri-Light数据集的未标记音频来获得半监督学习中最新的发展的最新发展，以获得自动语音识别的最新结果。更确切地说，我们使用使用WAV2VEC 2.0预训练的巨型构象模型进行了嘈杂的学生培训，并使用巨型构象模型进行了训练。通过这样做，我们能够在Librispeech测试/测试中获得1.4％/2.6％的单词率率（WERS），而目前的最新设备为1.7％/3.3％。

屏蔽语言模型（MLMS），如BERT和ROBERTA，在过去几年中彻底改变了自然语言理解领域。然而，现有的预先训练的MLMS通常输出令牌表示的各向异性分布，其占据整个表示空间的窄子集。这些令牌表示不理想，特别是对于要求不同令牌的判别语义含义的任务。在这项工作中，我们提出了TACL（令牌感知的对比学习），这是一种新的持续预训练方法，鼓励伯特来学习令牌陈述的各向同性和鉴别分布。TACL完全无监督，无需其他数据。我们在广泛的英语和中国基准上广泛地测试了我们的方法。结果表明，TACL通过原始BERT模型带来一致和显着的改进。此外，我们进行了详细的分析，以揭示我们方法的优点和内在运作。

本文研究了从预先训练的模型，尤其是蒙面自动编码器中提取知识的潜力。我们的方法很简单：除了优化掩盖输入的像素重建损失外，我们还将教师模型的中间特征图与学生模型的中间特征图之间的距离最小化。此设计导致一个计算高效的知识蒸馏框架，给定1）仅使用一个少量可见的补丁子集，2）（笨拙的）教师模型仅需要部分执行，\ ie，\ ie，在前几个中，向前传播输入层，用于获得中间特征图。与直接蒸馏微型模型相比，提炼预训练的模型显着改善了下游性能。例如，通过将知识从MAE预先训练的VIT-L提炼为VIT-B，我们的方法可实现84.0％的Imagenet Top-1精度，表现优于直接将微型VIT-L蒸馏的基线，降低1.2％。更有趣的是，我们的方法即使具有极高的掩盖率也可以从教师模型中进行鲁棒性蒸馏：例如，在蒸馏过程中仅可见十个斑块，我们的VIT-B具有竞争力的前1个Imagenet精度为83.6％，在95％的掩盖率中，只有十个斑块。 ;令人惊讶的是，它仍然可以通过仅四个可见斑（98％的掩盖率）积极训练来确保82.4％的Top-1 Imagenet精度。代码和模型可在https://github.com/ucsc-vlaa/dmae上公开获得。

Benefiting from masked visual modeling, self-supervised video representation learning has achieved remarkable progress. However, existing methods focus on learning representations from scratch through reconstructing low-level features like raw pixel RGB values. In this paper, we propose masked video distillation (MVD), a simple yet effective two-stage masked feature modeling framework for video representation learning: firstly we pretrain an image (or video) model by recovering low-level features of masked patches, then we use the resulting features as targets for masked feature modeling. For the choice of teacher models, we observe that students taught by video teachers perform better on temporally-heavy video tasks, while image teachers transfer stronger spatial representations for spatially-heavy video tasks. Visualization analysis also indicates different teachers produce different learned patterns for students. Motivated by this observation, to leverage the advantage of different teachers, we design a spatial-temporal co-teaching method for MVD. Specifically, we distill student models from both video teachers and image teachers by masked feature modeling. Extensive experimental results demonstrate that video transformers pretrained with spatial-temporal co-teaching outperform models distilled with a single teacher on a multitude of video datasets. Our MVD with vanilla ViT achieves state-of-the-art performance compared with previous supervised or self-supervised methods on several challenging video downstream tasks. For example, with the ViT-Large model, our MVD achieves 86.4% and 75.9% Top-1 accuracy on Kinetics-400 and Something-Something-v2, outperforming VideoMAE by 1.2% and 1.6% respectively. Code will be available at \url{https://github.com/ruiwang2021/mvd}.

最近的蒙版图像建模（MIM）在自我监督学习（SSL）中受到了很多关注，该学习要求目标模型恢复输入图像的掩盖部分。尽管基于MIM的预训练方法在转移到许多下游任务时达到了新的最新性能，但可视化表明，与基于基于对比性学习预训练相比，学习的表示形式不可分割，尤其是相比。这激发了我们思考MIM预培训表示的线性可分离性是否可以进一步改善，从而改善了训练的性能。由于MIM和对比度学习倾向于利用不同的数据增强和培训策略，因此将这两个借口任务结合起来并不是微不足道的。在这项工作中，我们提出了一个新颖而灵活的预训练框架，名为Mimco，该框架通过两阶段的预培训结合了MIM和对比度学习。具体而言，MIMCO将预先训练的对比学习模型作为教师模型，并通过两种类型的学习目标进行了预培训：贴片级和图像级的重建损失。关于下游任务的广泛转移实验证明了我们的MIMCO预训练框架的出色表现。以VIT-S为例，当使用预先训练的MoCov3-Vit-S作为教师模型时，Mimco只需要100个时期的预训练时期即可达到Imagenet-1K上的82.53％Top-1 FineTuning精度，这表现优于表现最先进的自我监督学习对手。

自我监督学习（SSL）被视为一种非常有前途的方法，对于下游任务的几个语音，高性能。由于SSL模型的参数通常是如此之大，以至于训练和推理需要大量的内存和计算成本，因此希望通过应用诸如知识蒸馏（KD）等压缩方法来生成紧凑的SSL模型，而无需显着性能降解。尽管KD方法能够缩小SSL模型结构的深度和/或宽度，但几乎没有研究如何改变深度和宽度对小脚印模型的内部表示。本文提供了一项解决问题的经验研究。我们在改变结构和KD方法的同时研究了Superb的性能，以保持参数恒定的数量；这使我们能够分析通过改变模型体系结构引入的表示的贡献。实验表明，一定深度对于准确地求解面向内容的任务（例如自动语音识别）至关重要，而在几个面向讲话者的任务上（例如，说话者的身份），必须进行一定宽度对于实现高性能。基于这些观察结果，我们确定了与以前的研究相比，具有更好性能的更高压模型。

当前的领先错误发音检测和诊断（MDD）系统通过端到端音素识别实现有希望的性能。这种端到端解决方案的一个挑战是在自然L2语音上缺乏人类注销的音素。在这项工作中，我们通过伪标记（PL）程序利用未标记的L2语音，并扩展基于预先训练的自我监督学习（SSL）模型的微调方法。具体而言，我们使用WAV2VEC 2.0作为我们的SSL模型，并使用原始标记的L2语音样本以及创建的伪标记的L2语音样本进行微调。我们的伪标签是动态的，是由在线模型的合奏生成的，这确保了我们的模型对伪标签的噪声具有强大的功能。我们表明，使用伪标签进行微调可实现5.35％的音素错误率降低和2.48％的MDD F1得分在仅标签样本的基线基线。提出的PL方法还显示出优于常规的离线PL方法。与最先进的MDD系统相比，我们的MDD解决方案会产生更准确，一致的语音误差诊断。此外，我们对单独的UTD-4ACCENTS数据集进行了开放测试，在该数据集中，我们的系统识别输出基于重音和清晰度，与人类感知有着密切的相关性。