End-to-end speaker diarization for an unknown number of speakers is addressed
in this paper. Recently proposed end-to-end speaker diarization outperformed
conventional clustering-based speaker diarization, but it has one drawback: it
is less flexible in terms of the number of speakers. This paper proposes a
method for encoder-decoder based attractor calculation (EDA), which first
generates a flexible number of attractors from a speech embedding sequence.
Then, the generated multiple attractors are multiplied by the speech embedding
sequence to produce the same number of speaker activities. The speech embedding
sequence is extracted using the conventional self-attentive end-to-end neural
speaker diarization (SA-EEND) network. In a two-speaker condition, our method
achieved a 2.69 % diarization error rate (DER) on simulated mixtures and a 8.07
% DER on the two-speaker subset of CALLHOME, while vanilla SA-EEND attained
4.56 % and 9.54 %, respectively. In unknown numbers of speakers conditions, our
method attained a 15.29 % DER on CALLHOME, while the x-vector-based clustering
method achieved a 19.43 % DER.

该论文提出了一种基于编码器 - 解码器的吸引因子计算方法，该方法可灵活地生成不同数量的吸引因子，用以支持不同数量的说话人，并利用传统的自我注意力端到端神经说话人分离网络（SA-EEND）来提取讲话嵌入序列和生成说话人活动，实验结果表明，相对于传统的基于聚类的说话人分离模型和端到端模型，该方法分别在模拟的双说话人条件和未知说话人数量的条件下都取得了更好的说话人分离性能。

基于编码器 - 解码器吸引子的未知数量说话者端到端分离

End-to-End Speaker Diarization for an Unknown Number of Speakers with  Encoder-Decoder Based Attractors

Asynchronous events on the continuous time domain, e.g., social media actions
and stock transactions, occur frequently in the world. The ability to recognize
occurrence patterns of event sequences is crucial to predict which typeof
events will happen next and when. A de facto standard mathematical framework to
do this is the Hawkes process. In order to enhance expressivity of multivariate
Hawkes processes, conventional statistical methods and deep recurrent networks
have been employed to modify its intensity function. The former is highly
interpretable and requires small size of training data but relies on correct
model design while the latter has less dependency on prior knowledge and is
more powerful in capturing complicated patterns. We leverage pros and cons of
these models and propose a self-attentive Hawkes process(SAHP). The proposed
method adapts self-attention to fit the intensity function of Hawkes processes.
This design has two benefits:(1) compared with conventional statistical
methods, the SAHP is more powerful to identify complicated dependency
relationships between temporal events; (2)compared with deep recurrent
networks, the self-attention mechanism is able to capture longer historical
information, and is more interpretable because the learnt attention weight
tensor shows contributions of each historical event. Experiments on four
real-world datasets demonstrate the effectiveness of the proposed method.

提出了一种利用 self-attention 机制进行 intensity function 拟合的 self-attentive Hawkes process 方法，相较于传统的统计方法和深度循环神经网络，该方法能更好地识别时间事件之间的复杂依赖关系，并且能够捕捉更长的历史信息，可以针对多变量事件序列进行有效的复杂模式预测。

自我关注 Hawkes 过程

Self-Attentive Hawkes Processes

Recurrent neural networks (RNNs) sequentially process data by updating their
state with each new data point, and have long been the de facto choice for
sequence modeling tasks. However, their inherently sequential computation makes
them slow to train. Feed-forward and convolutional architectures have recently
been shown to achieve superior results on some sequence modeling tasks such as
machine translation, with the added advantage that they concurrently process
all inputs in the sequence, leading to easy parallelization and faster training
times. Despite these successes, however, popular feed-forward sequence models
like the Transformer fail to generalize in many simple tasks that recurrent
models handle with ease, e.g. copying strings or even simple logical inference
when the string or formula lengths exceed those observed at training time. We
propose the Universal Transformer (UT), a parallel-in-time self-attentive
recurrent sequence model which can be cast as a generalization of the
Transformer model and which addresses these issues. UTs combine the
parallelizability and global receptive field of feed-forward sequence models
like the Transformer with the recurrent inductive bias of RNNs. We also add a
dynamic per-position halting mechanism and find that it improves accuracy on
several tasks. In contrast to the standard Transformer, under certain
assumptions, UTs can be shown to be Turing-complete. Our experiments show that
UTs outperform standard Transformers on a wide range of algorithmic and
language understanding tasks, including the challenging LAMBADA language
modeling task where UTs achieve a new state of the art, and machine translation
where UTs achieve a 0.9 BLEU improvement over Transformers on the WMT14 En-De
dataset.

Universal Transformer 是一种并行自我关注循环序列模型，它将前馈序列模型（例如 Transformer）的并行性和全局感受野与递归的归纳偏差相结合，解决了在某些简单任务中标准 Transformer 失败的问题，并取得诸多算法和语言理解任务上优越的表现。