CEEMDAN-IPSO-LSTM: A Novel Model for Short-Term Passenger Flow Prediction in Urban Rail Transit Systems.

Urban rail transit (URT) is a key mode of public transport, which serves for greatest user demand. Short-term passenger flow prediction aims to improve management validity and avoid extravagance of public transport resources. In order to anticipate passenger flow for URT, managing nonlinearity, correlation, and periodicity of data series in a single model is difficult. This paper offers a short-term passenger flow prediction combination model based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and long-short term memory neural network (LSTM) in order to more accurately anticipate the short-period passenger flow of URT. In the meantime, the hyperparameters of LSTM were calculated using the improved particle swarm optimization (IPSO). First, CEEMDAN-IPSO-LSTM model performed the CEEMDAN decomposition of passenger flow data and obtained uncoupled intrinsic mode functions and a residual sequence after removing noisy data. Second, we built a CEEMDAN-IPSO-LSTM passenger flow prediction model for each decomposed component and extracted prediction values. Third, the experimental results showed that compared with the single LSTM model, CEEMDAN-IPSO-LSTM model reduced by 40 persons/35 persons, 44 persons/35 persons, 37 persons/31 persons, and 46.89%/35.1% in SD, RMSE, MAE, and MAPE, and increase by 2.32%/3.63% and 2.19%/1.67% in R and R2, respectively. This model can reduce the risks of public health security due to excessive crowding of passengers (especially in the period of COVID-19), as well as reduce the negative impact on the environment through the optimization of traffic flows, and develop low-carbon transportation.

Hardware Inspired Neural Network for Efficient Time-Resolved Biomedical Imaging.

Convolutional neural networks (CNN) have revealed exceptional performance for fluorescence lifetime imaging (FLIM). However, redundant parameters and complicated topologies make it challenging to implement such networks on embedded hardware to achieve real-time processing. We report a lightweight, quantized neural architecture that can offer fast FLIM imaging. The forward-propagation is significantly simplified by replacing matrix multiplications in each convolution layer with additions and data quantization using a low bit-width. We first used synthetic 3-D lifetime data with given lifetime ranges and photon counts to assure correct average lifetimes can be obtained. Afterwards, human prostatic cancer cells incubated with gold nanoprobes were utilized to validate the feasibility of the network for real-world data. The quantized network yielded a 37.8% compression ratio without performance degradation. Clinical relevance - This neural network can be applied to diagnose cancer early based on fluorescence lifetime in a non-invasive way. This approach brings high accuracy and accelerates diagnostic processes for clinicians who are not experts in biomedical signal processing.

Fast Analysis of Time-Domain Fluorescence Lifetime Imaging via Extreme Learning Machine.

We present a fast and accurate analytical method for fluorescence lifetime imaging microscopy (FLIM), using the extreme learning machine (ELM). We used extensive metrics to evaluate ELM and existing algorithms. First, we compared these algorithms using synthetic datasets. The results indicate that ELM can obtain higher fidelity, even in low-photon conditions. Afterwards, we used ELM to retrieve lifetime components from human prostate cancer cells loaded with gold nanosensors, showing that ELM also outperforms the iterative fitting and non-fitting algorithms. By comparing ELM with a computational efficient neural network, ELM achieves comparable accuracy with less training and inference time. As there is no back-propagation process for ELM during the training phase, the training speed is much higher than existing neural network approaches. The proposed strategy is promising for edge computing with online training.

Near-Ultraviolet Indoor Black Light-Harvesting Perovskite Solar Cells.

Indoor light-energy-harvesting solar cells have long-standing history with perovskite solar cells (PSCs) recently emerging as potential candidates with high power conversion efficiencies (PCEs). However, almost all of the reported studies on indoor light-harvesting solar cells utilize white light in the visible wavelength. Low wavelength near-ultraviolet (UV) lights used under indoor environments are not given attention despite their high photon energy. In this study, perovskite solar cells have been investigated for the first time for harvesting energy from a commercially available near-UV (UV-A) indoor LED light (395-400 nm). Also called black lights, these near-UV lights are commonly used for decoration (e.g., in bars, pubs, aquariums, parties, clubs, body art studios, neon lights, and Christmas and Halloween decorations). The optimized perovskite solar cells with the n-i-p architecture using the CH3NH3PbI3 absorber were fabricated and characterized under different illumination intensities of near-UV indoor LEDs. The champion devices delivered a PCE and power output of 20.63% and 775.86 µW/cm2, respectively, when measured under UV illumination of 3.76 mW/cm2. The devices retained 84.10% of their initial PCE when aged under near-UV light for 24 h. The effects of UV exposure on the device performance have been comprehensively characterized. Furthermore, UV-stable solar cells fabricated with a modified electron transport layer retained 95.53% of its initial PCE after 24 h UV exposure. The champion devices delivered enhanced PCE and power output of 26.19% and 991.21 µW/cm2, respectively, when measured under UV illumination of 3.76 mW/cm2. This work opens up a novel direction for energy harvesting from near-UV indoor light sources for applications in microwatt-powered electronics such as internet of things sensors.