In an autonomous driving system (ADS), the perception module is crucial to driving safety and efficiency. Unfortunately, the perception in today's ADS remains oblivious to driving decisions, contrasting to how humans drive. Our idea is to refactor ADS so that (1) the ADS guides its perception with the driving knowledge in situ; (2) the perception differentiates between awareness and attention. We propose a system called AnA with three novel mechanisms: (1) a query interface for the planning to express its interest in perception; (2) a query executor that maps queries to an optimal set of perception tasks; (3) a monitor for handling abnormal task executions with driving knowledge. On challenging driving benchmarks, AnA outperforms competitive baselines: it responds to adversarial events timely, reducing collisions by 2x; it reduces compute usage by 44% without compromising driving safety. We attribute AnA's efficacy to its attentive driving, a human-like behavior that improves resource proportionality.

Transformer-based speech and language models deliver high-quality transcription and context-aware responses but require significant resources, complicating on-device deployment. Our work aims to build efficient mobile systems for real-time, accurate on-device model processing through system- and runtime-level innovations, thus eliminating cloud dependency and enhancing privacy.

Speech foundation models, such as OpenAI's Whisper, become the state of the art in speech understanding due to their strong accuracy and generalizability. Yet, their applications are mostly limited to processing pre-recorded speech, whereas processing of streaming speech, in particular doing it efficiently, remains rudimentary.

Speech foundation models, such as OpenAI's Whisper, become the state of the art in speech understanding due to their strong accuracy and generalizability. Yet, their applications are mostly limited to processing pre-recorded speech, whereas processing of streaming speech, in particular doing it efficiently, remains rudimentary. Behind this inefficiency are multiple fundamental reasons: (1) speech foundation models are trained to process long, fixed-length voice inputs (often 30 seconds); (2) encoding each voice input requires encoding as many as 1,500 tokens with tens of transformer layers; (3) decoding each output entails an irregular, complex beam search. As such, streaming speech processing on resource-constrained client devices is more expensive than other AI tasks, e.g., text generation.

Proto is a new instructional OS that runs on commodity, portable hardware. It showcases modern features, including per-app address spaces, threading, commodity filesystems, USB, DMA, multicore support, self-hosted debugging, and a window manager. It supports rich applications such as 2D/3D games, music and video players, and a blockchain miner. Unlike traditional instructional systems, Proto emphasizes engaging, media-rich apps that go beyond basic terminal programs. Our method breaks down a full-featured OS into a set of incremental, self-contained prototypes. Each prototype introduces a minimal set of OS mechanisms, driven by the needs of specific apps. The construction process then progressively enables these apps by bringing up one mechanism at a time.

Modern speech understanding (SU) runs a sophisticated pipeline: ingesting streaming voice input, the pipeline executes encoder-decoder based deep neural networks repeatedly; by doing so, the pipeline generates tentative outputs (called hypotheses), and periodically scores the hypotheses.

SUMMARY: Recent innovations in single-cell chromatin accessibility sequencing (scCAS) have revolutionized the characterization of epigenomic heterogeneity. Estimation of the number of cell types is a crucial step for downstream analyses and biological implications. However, efforts to perform estimation specifically for scCAS data are limited. Here, we propose ASTER, an ensemble learning-based tool for accurately estimating the number of cell types in scCAS data. ASTER outperformed baseline methods in systematic evaluation on 27 datasets of various protocols, sizes, numbers of cell types, degrees of cell-type imbalance, cell states and qualities, providing valuable guidance for scCAS data analysis. AVAILABILITY AND IMPLEMENTATION: ASTER along with detailed documentation is freely accessible at https://aster.readthedocs.io/ under the MIT License. It can be seamlessly integrated into existing scCAS analysis workflows. The source code is available at https://github.com/biox-nku/aster. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Digital beamforming (DBF) can provide high-gain narrow-beam scanning reception while transmitting wide-beam signals, which greatly improves the signal-to-noise ratio (SNR) of the corresponding systems. It is an effective technique for synthetic aperture radar (SAR) to obtain high-resolution wide-swath (HRWS) imaging capability. However, elevation changes in mountain area will lead to beam-pointing mismatch problems when using the ideal sphere model to calculate the beamforming weighting vector. As a result, the loss of receive gain and the deterioration of the SNR will occur. To solve this problem, adaptive DBF (ADBF) methods based on spectral estimation are typically used, such as Capon and MUSIC. However, the computational complexity of spectral estimation method is high, which is not conducive to on-satellite real-time processing. Therefore, a low complexity ADBF method based on beam-space phase-center cross correlation is proposed. In this method, the whole array is divided into several subarrays, and multiple phase centers are formed by beamforming so that the angle of arrival (AOA) of the signal source can be accurately estimated. Then, the weighted vector of the received beam is updated to mitigate the loss of receiver gain. The simulation results and airborne measured data validate the effectiveness of the proposed method. Compared with methods based on Capon and MUSIC, the proposed method can decrease the computational complexity without reducing the processing accuracy, thus providing a basis for the real-time processing of spaceborne DBF-SAR signals in the future.

The multiple elevation beam (MEB) mode is an effective technique for enhancing imaging width in spaceborne synthetic aperture radar (SAR) systems. This mode combines intra-pulse beam-steering during transmitting and digital beamforming (DBF) during receiving. By sequentially illuminating the far sub-swath followed by the near sub-swath, echoes from different sub-swaths can reach the antenna at the same time and overlap each other in the receiving window. To separate the overlapping echoes, the linear constrained minimum variance (LCMV) beamformer has been used, which is a single-null constrained beamformer. Additionally, a multi-null constraint beamformer has also been proposed on this basis. However, these two methods are insufficient for effectively separating the overlapping echoes when there is a significant energy difference between different sub-beams signals. To solve this problem, an advanced multi-null constrained beamformer with deepened nulls is proposed. Compared with other methods, the proposed method can flexibly adjust the width and depth of the nulls. The simulation results demonstrate that the proposed method can enhance echo separation quality. And the experimental results verify the effectiveness of the proposed method. All the results indicate that the proposed method is helpful to improve high-resolution and wide-swath imaging performance of future spaceborne SAR systems.