Sparsely Overlapped Speech Training in the Time Domain: Joint Learning of Target Speech Separation and Personal VAD Benefits

Target speech separation is the process of filtering a certain speaker's voice out of speech mixtures according to the additional speaker identity information provided. Recent works have made considerable improvement by processing signals in the time domain directly. The majority of them take fully overlapped speech mixtures for training. However, since most real-life conversations occur randomly and are sparsely overlapped, we argue that training with different overlap ratio data benefits. To do so, an unavoidable problem is that the popularly used SI-SNR loss has no definition for silent sources. This paper proposes the weighted SI-SNR loss, together with the joint learning of target speech separation and personal VAD. The weighted SI-SNR loss imposes a weight factor that is proportional to the target speaker's duration and returns zero when the target speaker is absent. Meanwhile, the personal VAD generates masks and sets non-target speech to silence. Experiments show that our proposed method outperforms the baseline by 1.73 dB in terms of SDR on fully overlapped speech, as well as by 4.17 dB and 0.9 dB on sparsely overlapped speech of clean and noisy conditions. Besides, with slight degradation in performance, our model could reduce the time costs in inference.

EF-BV: A Unified Theory of Error Feedback and Variance Reduction Mechanisms for Biased and Unbiased Compression in Distributed Optimization

In distributed or federated optimization and learning, communication between the different computing units is often the bottleneck, and gradient compression is a widely used technique for reducing the number of bits sent within each communication round of iterative methods. There are two classes of compression operators and separate algorithms making use of them. In the case of unbiased random compressors with bounded variance (e.g., rand-k), the DIANA algorithm of Mishchenko et al. [2019], which implements a variance reduction technique for handling the variance introduced by compression, is the current state of the art. In the case of biased and contractive compressors (e.g., top-k), the EF21 algorithm of Richt\'arik et al. [2021], which implements an error-feedback mechanism for handling the error introduced by compression, is the current state of the art. These two classes of compression schemes and algorithms are distinct, with different analyses and proof techniques. In this paper, we unify them into a single framework and propose a new algorithm, recovering DIANA and EF21 as particular cases. We prove linear convergence under certain conditions. Our general approach works with a new, larger class of compressors, which includes unbiased and biased compressors as particular cases, and has two parameters, the bias and the variance. These gives a finer control and allows us to inherit the best of the two worlds: biased compressors, whose good performance in practice is recognized, can be used. And independent randomness at the compressors allows to mitigate the effects of compression, with the convergence rate improving when the number of parallel workers is large. This is the first time that an algorithm with all these features is proposed. Our approach takes a step towards better understanding of two so-far distinct worlds of communication-efficient distributed learning.

Efficient Multiscale Object-based Superpixel Framework

Superpixel segmentation can be used as an intermediary step in many applications, often to improve object delineation and reduce computer workload. However, classical methods do not incorporate information about the desired object. Deep-learning-based approaches consider object information, but their delineation performance depends on data annotation. Additionally, the computational time of object-based methods is usually much higher than desired. In this work, we propose a novel superpixel framework, named Superpixels through Iterative CLEarcutting (SICLE), which exploits object information being able to generate a multiscale segmentation on-the-fly. SICLE starts off from seed oversampling and repeats optimal connectivity-based superpixel delineation and object-based seed removal until a desired number of superpixels is reached. It generalizes recent superpixel methods, surpassing them and other state-of-the-art approaches in efficiency and effectiveness according to multiple delineation metrics.

Traffic Context Aware Data Augmentation for Rare Object Detection in Autonomous Driving

Detection of rare objects (e.g., traffic cones, traffic barrels and traffic warning triangles) is an important perception task to improve the safety of autonomous driving. Training of such models typically requires a large number of annotated data which is expensive and time consuming to obtain. To address the above problem, an emerging approach is to apply data augmentation to automatically generate cost-free training samples. In this work, we propose a systematic study on simple Copy-Paste data augmentation for rare object detection in autonomous driving. Specifically, local adaptive instance-level image transformation is introduced to generate realistic rare object masks from source domain to the target domain. Moreover, traffic scene context is utilized to guide the placement of masks of rare objects. To this end, our data augmentation generates training data with high quality and realistic characteristics by leveraging both local and global consistency. In addition, we build a new dataset named NM10k consisting 10k training images, 4k validation images and the corresponding labels with a diverse range of scenarios in autonomous driving. Experiments on NM10k show that our method achieves promising results on rare object detection. We also present a thorough study to illustrate the effectiveness of our local-adaptive and global constraints based Copy-Paste data augmentation for rare object detection. The data, development kit and more information of NM10k dataset are available online at: \url{https://nullmax-vision.github.io}.

Exact Indexing of Time Series under Dynamic Time Warping

Dynamic time warping (DTW) is a robust similarity measure of time series. However, it does not satisfy triangular inequality and has high computational complexity, severely limiting its applications in similarity search on large-scale datasets. Usually, we resort to lower bounding distances to speed up similarity search under DTW. Unfortunately, there is still a lack of an effective lower bounding distance that can measure unequal-length time series and has desirable tightness. In the paper, we propose a novel lower bounding distance LB_Keogh+, which is a seamless combination of sequence extension and LB_Keogh. It can be used for unequal-length sequences and has low computational complexity. Besides, LB_Keogh+ can extend sequences to an arbitrary suitable length, without significantly reducing tightness. Next, based on LB_Keogh+, an exact index of time series under DTW is devised. Then, we introduce several theorems and complete the relevant proofs to guarantee no false dismissals in our similarity search. Finally, extensive experiments are conducted on real-world datasets. Experimental results indicate that our proposed method can perform similarity search of unequal-length sequences with high tightness and good pruning power.

Single-pixel imaging based on weight sort of the Hadamard basis

Single-pixel imaging (SPI) is very popular in subsampling applications, but the random measurement matrices it typically uses will lead to measurement blindness as well as difficulties in calculation and storage, and will also limit the further reduction in sampling rate. The deterministic Hadamard basis has become an alternative choice due to its orthogonality and structural characteristics. There is evidence that sorting the Hadamard basis is beneficial to further reduce the sampling rate, thus many orderings have emerged, but their relations remain unclear and lack a unified theory. Given this, here we specially propose a concept named selection history, which can record the Hadamard spatial folding process, and build a model based on it to reveal the formation mechanisms of different orderings and to deduce the mutual conversion relationship among them. Then, a weight ordering of the Hadamard basis is proposed. Both numerical simulation and experimental results have demonstrated that with this weight sort technique, the sampling rate, reconstruction time and matrix memory consumption are greatly reduced in comparison to traditional sorting methods. Therefore, we believe that this method may pave the way for real-time single-pixel imaging.

Real-time Travel Time Estimation Using Matrix Factorization

Estimating the travel time of any route is of great importance for trip planners, traffic operators, online taxi dispatching and ride-sharing platforms, and navigation provider systems. With the advance of technology, many traveling cars, including online taxi dispatch systems' vehicles are equipped with Global Positioning System (GPS) devices that can report the location of the vehicle every few seconds. This paper uses GPS data and the Matrix Factorization techniques to estimate the travel times on all road segments and time intervals simultaneously. We aggregate GPS data into a matrix, where each cell of the original matrix contains the average vehicle speed for a segment and a specific time interval. One of the problems with this matrix is its high sparsity. We use Alternating Least Squares (ALS) method along with a regularization term to factorize the matrix. Since this approach can solve the sparsity problem that arises from the absence of cars in many road segments in a specific time interval, matrix factorization is suitable for estimating the travel time. Our comprehensive evaluation results using real data provided by one of the largest online taxi dispatching systems in Iran, shows the strength of our proposed method.

Fractional Vegetation Cover Estimation using Hough Lines and Linear Iterative Clustering

A common requirement of plant breeding programs across the country is companion planting -- growing different species of plants in close proximity so they can mutually benefit each other. However, the determination of companion plants requires meticulous monitoring of plant growth. The technique of ocular monitoring is often laborious and error prone. The availability of image processing techniques can be used to address the challenge of plant growth monitoring and provide robust solutions that assist plant scientists to identify companion plants. This paper presents a new image processing algorithm to determine the amount of vegetation cover present in a given area, called fractional vegetation cover. The proposed technique draws inspiration from the trusted Daubenmire method for vegetation cover estimation and expands upon it. Briefly, the idea is to estimate vegetation cover from images containing multiple rows of plant species growing in close proximity separated by a multi-segment PVC frame of known size. The proposed algorithm applies a Hough Transform and Simple Linear Iterative Clustering (SLIC) to estimate the amount of vegetation cover within each segment of the PVC frame. The analysis when repeated over images captured at regular intervals of time provides crucial insights into plant growth. As a means of comparison, the proposed algorithm is compared with SamplePoint and Canopeo, two trusted applications used for vegetation cover estimation. The comparison shows a 99% similarity with both SamplePoint and Canopeo demonstrating the accuracy and feasibility of the algorithm for fractional vegetation cover estimation.

A Survey on Semi-parametric Machine Learning Technique for Time Series Forecasting

Artificial Intelligence (AI) has recently shown its capabilities for almost every field of life. Machine Learning, which is a subset of AI, is a `HOT' topic for researchers. Machine Learning outperforms other classical forecasting techniques in almost all-natural applications. It is a crucial part of modern research. As per this statement, Modern Machine Learning algorithms are hungry for big data. Due to the small datasets, the researchers may not prefer to use Machine Learning algorithms. To tackle this issue, the main purpose of this survey is to illustrate, demonstrate related studies for significance of a semi-parametric Machine Learning framework called Grey Machine Learning (GML). This kind of framework is capable of handling large datasets as well as small datasets for time series forecasting likely outcomes. This survey presents a comprehensive overview of the existing semi-parametric machine learning techniques for time series forecasting. In this paper, a primer survey on the GML framework is provided for researchers. To allow an in-depth understanding for the readers, a brief description of Machine Learning, as well as various forms of conventional grey forecasting models are discussed. Moreover, a brief description on the importance of GML framework is presented.

Flowformer: Linearizing Transformers with Conservation Flows

Transformers based on the attention mechanism have achieved impressive success in various areas. However, the attention mechanism has a quadratic complexity, significantly impeding Transformers from dealing with numerous tokens and scaling up to bigger models. Previous methods mainly utilize the similarity decomposition and the associativity of matrix multiplication to devise linear-time attention mechanisms. They avoid degeneration of attention to a trivial distribution by reintroducing inductive biases such as the locality, thereby at the expense of model generality and expressiveness. In this paper, we linearize Transformers free from specific inductive biases based on the flow network theory. We cast attention as the information flow aggregated from the sources (values) to the sinks (results) through the learned flow capacities (attentions). Within this framework, we apply the property of flow conservation with attention and propose the Flow-Attention mechanism of linear complexity. By respectively conserving the incoming flow of sinks for source competition and the outgoing flow of sources for sink allocation, Flow-Attention inherently generates informative attentions without using specific inductive biases. Empowered by the Flow-Attention, Flowformer yields strong performance in linear time for wide areas, including long sequence, time series, vision, natural language, and reinforcement learning.