Prefix-tuning, or more generally continuous prompt tuning, has become an essential paradigm of parameter-efficient transfer learning. Using a large pre-trained language model (PLM), prefix-tuning can obtain strong performance by training only a small portion of parameters. In this paper, we propose to understand and further develop prefix-tuning through the kernel lens. Specifically, we make an analogy between \textit{prefixes} and \textit{inducing variables} in kernel methods and hypothesize that \textit{prefixes} serving as \textit{inducing variables} would improve their overall mechanism. From the kernel estimator perspective, we suggest a new variant of prefix-tuning -- \textit{inducer-tuning}, which shares the exact mechanism as prefix-tuning while leveraging the residual form found in adapter-tuning. This mitigates the initialization issue in prefix-tuning. Through comprehensive empirical experiments on natural language understanding and generation tasks, we demonstrate that inducer-tuning can close the performance gap between prefix-tuning and fine-tuning.
The conversation scenario is one of the most important and most challenging scenarios for speech processing technologies because people in conversation respond to each other in a casual style. Detecting the speech activities of each person in a conversation is vital to downstream tasks, like natural language processing, machine translation, etc. People refer to the detection technology of "who speak when" as speaker diarization (SD). Traditionally, diarization error rate (DER) has been used as the standard evaluation metric of SD systems for a long time. However, DER fails to give enough importance to short conversational phrases, which are short but important on the semantic level. Also, a carefully and accurately manually-annotated testing dataset suitable for evaluating the conversational SD technologies is still unavailable in the speech community. In this paper, we design and describe the Conversational Short-phrases Speaker Diarization (CSSD) task, which consists of training and testing datasets, evaluation metric and baselines. In the dataset aspect, despite the previously open-sourced 180-hour conversational MagicData-RAMC dataset, we prepare an individual 20-hour conversational speech test dataset with carefully and artificially verified speakers timestamps annotations for the CSSD task. In the metric aspect, we design the new conversational DER (CDER) evaluation metric, which calculates the SD accuracy at the utterance level. In the baseline aspect, we adopt a commonly used method: Variational Bayes HMM x-vector system, as the baseline of the CSSD task. Our evaluation metric is publicly available at https://github.com/SpeechClub/CDER_Metric.
Housing quality is an essential proxy for regional wealth, security and health. Understanding the distribution of housing quality is crucial for unveiling rural development status and providing political proposals. However,present rural house quality data highly depends on a top-down, time-consuming survey at the national or provincial level but fails to unpack the housing quality at the village level. To fill the gap between accurately depicting rural housing quality conditions and deficient data,we collect massive rural images and invite users to assess their housing quality at scale. Furthermore, a deep learning framework is proposed to automatically and efficiently predict housing quality based on crowd-sourcing rural images.
Randomly pivoted Cholesky (RPCholesky) is a natural algorithm for computing a rank-k approximation of an N x N positive semidefinite (psd) matrix. RPCholesky can be implemented with just a few lines of code. It requires only (k+1)N entry evaluations and O(k^2 N) additional arithmetic operations. This paper offers the first serious investigation of its experimental and theoretical behavior. Empirically, RPCholesky matches or improves on the performance of alternative algorithms for low-rank psd approximation. Furthermore, RPCholesky provably achieves near-optimal approximation guarantees. The simplicity, effectiveness, and robustness of this algorithm strongly support its use in scientific computing and machine learning applications.
We propose a system to visualize the chirality of the protein in brains, which would be helpful to diagnose early neurological degenerative diseases in vivo. These neurological degenerative diseases often occur along with some mark proteins. By nanoparticle instilling and metamaterial technique, the chiral effect of the mark proteins is assumed to be manifest in microwave regime. Therefore, by detecting the transmission of cross-polarization, we could detect the chirality that rotates the microwave polarization angle. We developed a numerical method to simulate the electromagnetic response upon chiral (bi-isotropic) material. Then a numerical experiment was conduct with a numerical head phantom. A map of cross-polarized transmission magnitude can be reached by sweeping the antenna pair. The imaging results matches well with the distribution of chiral materials. It suggests that the proposed method would be capable of in vivo imaging of neurological degenerative disease using microwaves.
For online speaker diarization, samples arrive incrementally, and the overall distribution of the samples is invisible. Moreover, in most existing clustering-based methods, the training objective of the embedding extractor is not designed specially for clustering. To improve online speaker diarization performance, we propose a unified online clustering framework, which provides an interactive manner between embedding extractors and clustering algorithms. Specifically, the framework consists of two highly coupled parts: clustering-guided recurrent training (CGRT) and truncated beam searching clustering (TBSC). The CGRT introduces the clustering algorithm into the training process of embedding extractors, which could provide not only cluster-aware information for the embedding extractor, but also crucial parameters for the clustering process afterward. And with these parameters, which contain preliminary information of the metric space, the TBSC penalizes the probability score of each cluster, in order to output more accurate clustering results in online fashion with low latency. With the above innovations, our proposed online clustering system achieves 14.48\% DER with collar 0.25 at 2.5s latency on the AISHELL-4, while the DER of the offline agglomerative hierarchical clustering is 14.57\%.
To accelerate the training of graph convolutional networks (GCNs), many sampling-based methods have been developed for approximating the embedding aggregation. Among them, a layer-wise approach recursively performs importance sampling to select neighbors jointly for existing nodes in each layer. This paper revisits the approach from a matrix approximation perspective. We identify two issues in the existing layer-wise sampling methods: sub-optimal sampling probabilities and the approximation bias induced by sampling without replacement. We propose two remedies: new sampling probabilities and a debiasing algorithm, to address these issues, and provide the statistical analysis of the estimation variance. The improvements are demonstrated by extensive analyses and experiments on common benchmarks.
The massive amount of trainable parameters in the pre-trained language models (PLMs) makes them hard to be deployed to multiple downstream tasks. To address this issue, parameter-efficient transfer learning methods have been proposed to tune only a few parameters during fine-tuning while freezing the rest. This paper looks at existing methods along this line through the \textit{kernel lens}. Motivated by the connection between self-attention in transformer-based PLMs and kernel learning, we propose \textit{kernel-wise adapters}, namely \textit{Kernel-mix}, that utilize the kernel structure in self-attention to guide the assignment of the tunable parameters. These adapters use guidelines found in classical kernel learning and enable separate parameter tuning for each attention head. Our empirical results, over a diverse set of natural language generation and understanding tasks, show that our proposed adapters can attain or improve the strong performance of existing baselines.
This paper introduces a high-quality rich annotated Mandarin conversational (RAMC) speech dataset called MagicData-RAMC. The MagicData-RAMC corpus contains 180 hours of conversational speech data recorded from native speakers of Mandarin Chinese over mobile phones with a sampling rate of 16 kHz. The dialogs in MagicData-RAMC are classified into 15 diversified domains and tagged with topic labels, ranging from science and technology to ordinary life. Accurate transcription and precise speaker voice activity timestamps are manually labeled for each sample. Speakers' detailed information is also provided. As a Mandarin speech dataset designed for dialog scenarios with high quality and rich annotations, MagicData-RAMC enriches the data diversity in the Mandarin speech community and allows extensive research on a series of speech-related tasks, including automatic speech recognition, speaker diarization, topic detection, keyword search, text-to-speech, etc. We also conduct several relevant tasks and provide experimental results to help evaluate the dataset.