Abstract Gully morphology characteristics can be used effectively to describe the status of gully development. The Chabagou watershed, located in hilly‐gully region Loess Plateau China, was selected investigate morphological using a 3D laser scanning technique (LIDAR). Thirty‐one representative gullies at different watershed locations and orders were chosen quantitatively establish empirical equations for estimating volume based on length surface area. Images point cloud data 31 collected,...

The early detection of skin cancer substantially improves the five-year survival rate patients. It is often difficult to distinguish malignant tumors from images, even by expert dermatologists. Therefore, several classification methods dermatoscopic images have been proposed, but they found be inadequate or defective for detection, and require a large amount calculations. This study proposes an improved capsule network called FixCaps dermoscopic image classification. can obtain larger...

While neural networks are powerful function approximators, they suffer from catastrophic forgetting when the data distribution is not stationary. One particular formalism that studies learning under non-stationary provided by continual learning, where non-stationarity imposed a sequence of distinct tasks. Most methods in this space assume, however, knowledge task boundaries, and focus on alleviating forgetting. In work, we depart view move towards faster remembering -- i.e measuring how...

Classic reinforcement learning (RL) theories cannot explain human behavior in the absence of external reward or when environment changes. Here, we employ a deep sequential decision-making paradigm with sparse and abrupt environmental To participants these environments, show that RL need to include surprise novelty, each distinct role. While novelty drives exploration before first encounter reward, increases rate world-model as well model-free action-values. Even though is available for...

In many daily tasks, we make multiple decisions before reaching a goal. order to learn such sequences of decisions, mechanism link earlier actions later reward is necessary. Reinforcement learning (RL) theory suggests two classes algorithms solving this credit assignment problem: classic temporal-difference learning, receive information only after repetitions the task, whereas models with eligibility traces reinforce entire from single experience (one-shot). Here, show one-shot sequences. We...

Direction-of-arrival (DOA) estimation for incoherently distributed (ID) sources is essential in multipath wireless communication scenarios, yet it remains challenging due to the combined effects of angular spread and gain-phase uncertainties antenna arrays. This paper presents a two-stage sparse DOA framework, transitioning from partial calibration full potential, under generalized array manifold (GAM) framework. In first stage, coarse estimates are obtained by exploiting output subset...

Accurate State of Health (SOH) estimation lithium-ion batteries (LIBs) is critical for ensuring the safety electric vehicles and improving reliability battery management systems (BMS). However, use individual health features (HFs) selection hyperparameters can increase data processing burden on BMS reduce accuracy data-driven models. To address above issue, this paper proposes a novel SOH method based PSO–GWO–LSSVM prediction model with multi-dimensional feature extraction. comprehensively...

State-of-energy (SOE) estimation helps to enhance the safety of battery operation and predict vehicle range. However, voltage plateau LiFePO4 (LFP) presents a significant challenge for SOE estimation. Therefore, this paper introduces significantly varying mechanical force feature tackle flat curve in mid-SOE region. A fusion model that integrates bidirectional long short-term memory (BiLSTM) network, particle swarm optimization (PSO), Kalman filter (KF) algorithm is proposed The BiLSTM...

It is generally considered that working memory (WM) capacity limited and WM affects cognitive processes. Distractor filtering efficiency has been suggested to be an important factor in determining the visual (VWM) of individuals. In present study, we investigated whether training (FE) could improve VWM capacity, as measured by performance on change detection task (CDT) changes CDA (contralateral delay activity) different conditions, evaluated transfer effect FE verbal fluid intelligence,...

This paper presents a novel multimodal dataset for the analysis of Quality Experience(QoE) in emerging immersive multimedia technologies. In particular, perceived Sense Presence (SoP) induced by one-minute long video stimuli is explored with respect to content, quality, resolution, and sound reproduction annotated subjective scores. Furthermore, complementary acquired physiological signals, such as EEG, ECG, respiration carried out, aiming at an alternative evaluation human experience while...

Analog, unclocked, spiking neuromorphic microchips open new perspectives for implantable or wearable biosensors and biocontrollers, due to their low energy consumption heat dissipation. However, the challenges from a computational point of view are formidable. Here we outline our solutions realize reservoir computing paradigm on such hardware address combined problems bit resolution, device mismatch, approximate neuron models, timescale mismatch. The main contribution is scheme, called...

Current value-based multi-agent reinforcement learning methods optimize individual Q values to guide individuals' behaviours via centralized training with decentralized execution (CTDE). However, such expected, i.e., risk-neutral, value is not sufficient even CTDE due the randomness of rewards and uncertainty in environments, which causes failure these train coordinating agents complex environments. To address issues, we propose RMIX, a novel cooperative MARL method Conditional Value at Risk...

Humans successfully explore their environment to find 'extrinsic' rewards, even when exploration requires several intermediate reward-free decisions. It has been hypothesized that 'intrinsic' rewards such as novelty guide this exploration. However, different intrinsic lead strategies, some prone suboptimal attraction irrelevant stochastic stimuli, sometimes called the 'noisy TV problem.' Here, we ask whether humans show a similar stochasticity and, if so, which type of reward guides We...

Catastrophic interference has been a major roadblock in the research of continual learning. Here we propose variant back-propagation algorithm, "conceptor-aided back-prop" (CAB), which gradients are shielded by conceptors against degradation previously learned tasks. Conceptors have their origin reservoir computing, where they shown to overcome catastrophic forgetting. CAB extends these results deep feedforward networks. On disjoint MNIST task outperforms two other methods for coping with...

With the rise of online e-commerce platforms, more and customers prefer to shop online. To sell products, platforms introduce various modules recommend items with different properties such as huge discounts. A web page often consists independent modules. The ranking policies these are decided by teams optimized individually without cooperation, which might result in competition between Thus, global policy whole could be sub-optimal. In this paper, we propose a novel multi-agent cooperative...

The olfactory sense is strongly related to memory and emotional processes. Studies on the effects of odor perception from brain activity have been conducted by using different neuro-imaging techniques. In this paper, we analyse electroencephalography (EEG) 23 subjects during perceiving pleasant unpleasant stimuli. We describe construction functional connectivity networks measured most commonly used models. discuss network-based features connectivity, design classifiers applying network...

Highway guardrail survey is an important task of highway management. In order to improve the accuracy detection in complex background, Mask RCNN was introduced combination with image preprocessing algorithm, Resnet101 used as backbone network, feature pyramid network (FPN) structure for extraction, and regional proposal (RPN) generate proposals each map. The mask guardrails generated by realize segmentation guardrails. average precision 200 test images 94.38%, recall rate 93.8%, MIoU...

SLAM technology has developed very rapidly in recent years, and systems for general environment achieved excellent results. However, there are still some practical problems affecting the robustness accuracy of SLAM, including impact highly dynamic environments. In this paper, we propose a real-time system, named DOS-SLAM, which is specifically designed The based on ORB-SLAM2, adds two modules to achieve accurate environments, object segmentation with instance geometric constraint static...

Connectionist models such as neural networks suffer from catastrophic forgetting. In this work, we study problem the perspective of information theory and define forgetting increase description lengths previous data when they are compressed with a sequentially learned model. addition, show that continual learning approaches based on variational posterior approximation generative replay can be considered approximations to two prequential coding methods in compression, namely, Bayesian mixture...

Abstract Classic reinforcement learning (RL) theories cannot explain human behavior in response to changes the environment or absence of external reward. Here, we design a deep sequential decision-making paradigm with sparse reward and abrupt environmental changes. To participants these environments, show that RL need include surprise novelty, each distinct role. While novelty drives exploration before first encounter reward, increases rate world-model as well model-free action-values. Even...