Search

Hyperspectral image (HSI) contains abundant spatial and spectral information, making it highly valuable for unmixing. In this paper, we propose a Dual-Stream Attention Network (DSANet) for HSI unmixing. The endmembers and abundance of a pixel in HSI have high correlations with its adjacent pixels. Therefore, we adopt a "many to one" strategy to estimate the abundance of the central pixel. In addition, we adopt multiview spectral method, dividing spectral bands into multiple partitions with low correlations to estimate abundances. To aggregate the estimated abundances for complementary from the two branches, we design a cross-fusion attention network to enhance valuable information. Extensive experiments have been conducted on two real datasets, which demonstrate the effectiveness of our DSANet., Comment: Accepted by IEEE IGARSS 2024

Receiving timely and personalized feedback is essential for second-language learners, especially when human instructors are unavailable. This study explores the effectiveness of Large Language Models (LLMs), including both proprietary and open-source models, for Automated Essay Scoring (AES). Through extensive experiments with public and private datasets, we find that while LLMs do not surpass conventional state-of-the-art (SOTA) grading models in performance, they exhibit notable consistency, generalizability, and explainability. We propose an open-source LLM-based AES system, inspired by the dual-process theory. Our system offers accurate grading and high-quality feedback, at least comparable to that of fine-tuned proprietary LLMs, in addition to its ability to alleviate misgrading. Furthermore, we conduct human-AI co-grading experiments with both novice and expert graders. We find that our system not only automates the grading process but also enhances the performance and efficiency of human graders, particularly for essays where the model has lower confidence. These results highlight the potential of LLMs to facilitate effective human-AI collaboration in the educational context, potentially transforming learning experiences through AI-generated feedback.

Over the past decade, the blockchain technology and its Bitcoin cryptocurrency have received considerable attention. Bitcoin has experienced significant price swings in daily and long-term valuations. In this paper, we propose a partial differential equation (PDE) model on the bitcoin transaction network for predicting bitcoin price. Through analysis of bitcoin subgraphs or chainlets, the PDE model captures the influence of transaction patterns on bitcoin price over time and combines the effect of all chainlet clusters. In addition, Google Trends Index is incorporated to the PDE model to reflect the effect of bitcoin market sentiment. The experiment shows that the average accuracy of daily bitcoin price prediction is 0.82 for 362 consecutive days in 2017. The results demonstrate the PDE model is capable of predicting bitcoin price. The paper is the first attempt to apply a PDE model to the bitcoin transaction network for predicting bitcoin price.

Accurate reporting and forecasting of PM2.5 concentration are important for improving public health. In this paper, we propose a daily prediction method of PM2.5 concentration by using data-driven ordinary differential equation (ODE) models. Specifically, based on the historical PM2.5 concentration, this method combines genetic programming and orthogonal least square method to evolve the ODE models, which describe the transport of PM2.5 and then uses the data-driven ODEs to predict the air quality in the future. Experiment results show that the ODE models obtain similar prediction results as the typical statistical model, and the prediction results from this method are relatively good. To our knowledge, this is the first attempt to evolve data-driven ODE models to study PM2.5 prediction.

Quorum sensing is the mechanism by which bacteria communicate and synchronize group behaviors. Quantitative information on parameters such as the copy number of particular quorum-sensing proteins should contribute strongly to understanding how the quorum-sensing network functions. Here we show that the copy number of the master regulator protein LuxR in Vibrio harveyi, can be determined in vivo by exploiting small-number fluctuations of the protein distribution when cells undergo division. When a cell divides, both its volume and LuxR protein copy number N are partitioned with slight asymmetries. We have measured the distribution functions describing the partitioning of the protein fluorescence and the cell volume. The fluorescence distribution is found to narrow systematically as the LuxR population increases while the volume partitioning is unchanged. Analyzing these changes statistically, we have determined that N = 80-135 dimers at low cell density and 575 dimers at high cell density. In addition, we have measured the static distribution of LuxR over a large (3,000) clonal population. Combining the static and time-lapse experiments, we determine the magnitude of the Fano factor of the distribution. This technique has broad applicability as a general, in vivo technique for measuring protein copy number and burst size., Comment: Main text 23 pages, 5 figures. Supporting material 19 pages, 7 figures. In new version, text revised, one figure reformatted

We have investigated the binding interaction between the bacteriophage lambda repressor CI and its target DNA using total internal reflection fluorescence microscopy. Large, step-wise changes in the intensity of the red fluorescent protein fused to CI were observed as it associated and dissociated from individually labeled single molecule DNA targets. The stochastic association and dissociation were characterized by Poisson statistics. Dark and bright intervals were measured for thousands of individual events. The exponential distribution of the intervals allowed direct determination of the association and dissociation rate constants, ka and kd respectively. We resolved in detail how ka and kd varied as a function of 3 control parameters, the DNA length L, the CI dimer concentration, and the binding affinity. Our results show that although interaction with non-operator DNA sequences are observable, CI binding to the operator site is not dependent on the length of flanking non-operator DNA., Comment: 34 pages, 10 figures, accepted by Biophysical Journal

We demonstrate a catalytic control method for speeding up the single-molecule DNA motor introduced by Li and Tan [Nano Lett. {\bf 2}, 315 (2002)]. A key rate-limiting barrier in the reaction part of the cycle is the tendency for the second fuel strand $B$ to fold into the chair-like configuration of the original motor strand $M$. This seriously impedes the restoration reaction. We have designed a catalytic strand to inhibit the folding of $B$. Introduction of the catalyst speeds up the restoration reaction by roughly a factor of 2. The catalyst shows robust behavior for more than one cycle. The experimental data can be understood with a model with intermediate products. This technique provides dynamic control of the restoration rate of the motor without affecting the straightening rate., Comment: 6 pages, 5 figures New figure (Fig. 6) and appendix added. Text revised