Quantiles are costly to compute exactly but can be efficiently estimated by quantile sketches. Extensive works on summarizing streaming data, such as KLL sketch, focus on minimizing the cost in memory to provide certain error guarantees. For the problem of quantile estimation of values in LSM-tree based stores, streaming methods have an expensive I/O cost linear to data size N. Since disk components (chunks and SSTables) in the LSM-tree are immutable once flushed, quantile sketches can be pre-computed as a type of statistics to reduce I/O cost and accelerate queries. Unfortunately, to provide deterministic additive εN error guarantees on queried data, all pre-computed deterministic sketches of queried chunks each with size N_c should provide εN_c error guarantee, resulting in no improvement in the linear I/O cost. In this study, we propose pre-computing randomized sketches which provide randomized additive error guarantees. Our major technical contributions include (1) randomized sketches for data chunks constructed in flush events, which are proved to be optimal and achieve an I/O cost proportional to √(N), (2) hierarchical randomized sketches for SSTables constructed in compaction events, that further improve the asymptotic I/O cost, (3) the KLL sketch summarizing proposed pre-computed sketches is proved to be more accurate than that summarizing streaming data, and proved to achieve sublinear I/O cost while achieving the same memory complexity as in the streaming settings. Extensive experiments on synthetic and real datasets demonstrate the superiority of the proposed techniques. The approach is deployed in an LSM-tree based time-series database Apache IoTDB.

It is time-consuming to render high-resolution images in applications such as video games and virtual reality, and thus super-resolution technologies become increasingly popular for real-time rendering. However, it is challenging to preserve sharp texture details, keep the temporal stability and avoid the ghosting artifacts in real-time super-resolution rendering. To address this issue, we introduce radiance demodulation to separate the rendered image or radiance into a lighting component and a material component, considering the fact that the light component is smoother than the rendered image so that the high-resolution material component with detailed textures can be easily obtained. We perform the super-resolution on the lighting component only and re-modulate it with the high-resolution material component to obtain the final super-resolution image with more texture details. A reliable warping module is proposed by explicitly marking the occluded regions to avoid the ghosting artifacts. To further enhance the temporal stability, we design a frame-recurrent neural network and a temporal loss to aggregate the previous and current frames, which can better capture the spatial-temporal consistency among reconstructed frames. As a result, our method is able to produce temporally stable results in real-time rendering with high-quality details, even in the challenging 4 × 4 super-resolution scenarios. Code is available at: https://github.com/Riga2/NSRD.

In this study, a multi-channel output synchronous pulse generator with adjustable phase and pulse width was designed to ensure phase coherence among multiple channels in a multi-channel synchronous pulse control system. The synchronous pulse generator with multiple outputs was constructed using discrete devices. The modulation of the output pulse width and the delay between channels was achieved by adjusting the resistance-capacitance characteristics of the RC circuit and the logic characteristics of the D flip-flop. The experimental results indicated that the synchronous pulse generator can achieve continuous and low-delay error-free triggering output for hundreds of cycles. At 10 V output, the modulation range was 210 ns ± 30 ns, with a maximum delay between channels of less than 60 ps and a rise time about 1 ns. This system can be used in various high-precision multi-channel synchronous control systems.

This article proposes a method for implementing sub-nanosecond fast edge pulse signals based on gallium nitride transistors combined with gate drivers. Sub-nanosecond fast edge pulse signals are commonly used as trigger signals in modern high-tech equipment. A comparison was made between two common fast edge pulse generation methods based on ultra high speed operational amplifiers and GaN HEMT combined driving. Based on the amplitude design requirements of practical applications, a method was proposed to achieve 5V --1 OV sub-nanosecond fast edge pulse signals through the combination of GaN HEMT and gate drivers. An analysis was conducted on the selection of GaN HEMT and gate drivers, and a simulation model based on the Cadence 16.6 Pspice AD platform was built based on the actual selection results. The influence of parameters such as dead time adjustment resistance, gate drive resistance, and bypass capacitance in the model on the output pulse signal was simulated and analyzed, and the optimal solution for achieving sub-nanosecond fast edge pulse signals of 5V ~ OV was ultimately determined.

Quantiles are fundamental statistics in various data science tasks, but costly to compute, e.g., by loading the entire data in memory for ranking. With limited memory space, prevalent in end devices or databases with heavy loads, it needs to scan the data in multiple passes. The idea is to gradually shrink the range of the queried quantile till it is small enough to fit in memory for ranking the result. Existing methods use deterministic sketches to determine the exact range of quantile, known as deterministic filter, which could be inefficient in range shrinking. In this study, we propose to shrink the ranges more aggressively, using randomized summaries such as KLL sketch. That is, with a high probability the quantile lies in a smaller range, namely probabilistic filter, determined by the randomized sketch. Specifically, we estimate the expected passes for determining the exact quantiles with probabilistic filters, and select a proper probability that can minimize the expected passes. Analyses show that our exact quantile determination method can terminate in P passes with 1-δ confidence, storing O(N 1/P logP-1/2P (1/δ)) items, close to the lower bound Ømega(N1/P) for a fixed δ. The approach has been deployed as a function in an LSM-tree based time-series database Apache IoTDB. Remarkably, the randomized sketches can be pre-computed for the immutable SSTables in LSM-tree. Moreover, multiple quantile queries could share the data passes for probabilistic filters in range estimation. Extensive experiments on real and synthetic datasets demonstrate the superiority of our proposal compared to the existing methods with deterministic filters. On average, our method takes 0.48 fewer passes and 18% of the time compared with the state-of-the-art deterministic sketch (GK sketch).

Median absolute deviation (MAD), the median of the absolute deviations from the median, has been found useful in various applications such as outlier detection. Together with median, MAD is more robust to abnormal data than mean and standard deviation (SD). Unfortunately, existing methods return only approximate MAD that may be far from the exact one, and thus mislead the downstream applications. Computing exact MAD is costly, however, especially in space, by storing the entire dataset in memory. In this paper, we propose COnstruction-REfinement Sketch (CORE-Sketch) for computing exact MAD. The idea is to construct some sketch within limited space, and gradually refine the sketch to find the MAD element, i.e., the element with distance to the median exactly equal to MAD. Mergeability and convergence of the method is analyzed, ensuring the correctness of the proposal and enabling parallel computation. Extensive experiments demonstrate that CORE-Sketch achieves significantly less space occupation compared to the aforesaid baseline of No-Sketch, and has time and space costs relatively comparable to the DD-Sketch method for approximate MAD.