Showing papers by "University of Electronic Science and Technology of China published in 2018"

Simple Baselines for Human Pose Estimation and Tracking

Abstract: There has been significant progress on pose estimation and increasing interests on pose tracking in recent years. At the same time, the overall algorithm and system complexity increases as well, making the algorithm analysis and comparison more difficult. This work provides simple and effective baseline methods. They are helpful for inspiring and evaluating new ideas for the field. State-of-the-art results are achieved on challenging benchmarks. The code will be available at https://github.com/leoxiaobin/pose.pytorch.

CdS-Based photocatalysts

Abstract: To solve the problem of the global energy shortage and the pollution of the environment, in recent years, semiconductor photocatalytic technology that converts solar energy into chemical fuel has been widely studied. Regarding semiconductor-based photocatalysts, CdS has attracted extensive attention due to its relatively narrow bandgap for visible-light response and sufficiently negative potential of the conduction band edge for the reduction of protons. Studies have shown that CdS-based photocatalysts possess excellent photocatalytic performance in terms of solar-fuel generation and environmental purification. This critical review presents the recent advances and progress in the design and synthesis of various CdS and CdS-based photocatalysts. The basic physical and chemical properties of CdS and the related growth mechanism have been briefly summarized. Moreover, the applications of CdS-based photocatalysts have been discussed such as in photocatalytic hydrogen production, reduction of CO2 to hydrocarbon fuels and degradation of pollutants. Finally, a brief perspective on the challenges and future directions for the development of CdS and CdS-based photocatalysts are also presented.

Additive Margin Softmax for Face Verification

Abstract: In this letter, we propose a conceptually simple and intuitive learning objective function, i.e., additive margin softmax, for face verification. In general, face verification tasks can be viewed as metric learning problems, even though lots of face verification models are trained in classification schemes. It is possible when a large-margin strategy is introduced into the classification model to encourage intraclass variance minimization. As one alternative, angular softmax has been proposed to incorporate the margin. In this letter, we introduce another kind of margin to the softmax loss function, which is more intuitive and interpretable. Experiments on LFW and MegaFace show that our algorithm performs better when the evaluation criteria are designed for very low false alarm rate.

A Survey on Learning to Hash

Abstract: Nearest neighbor search is a problem of finding the data points from the database such that the distances from them to the query point are the smallest. Learning to hash is one of the major solutions to this problem and has been widely studied recently. In this paper, we present a comprehensive survey of the learning to hash algorithms, categorize them according to the manners of preserving the similarities into: pairwise similarity preserving, multiwise similarity preserving, implicit similarity preserving, as well as quantization, and discuss their relations. We separate quantization from pairwise similarity preserving as the objective function is very different though quantization, as we show, can be derived from preserving the pairwise similarities. In addition, we present the evaluation protocols, and the general performance analysis, and point out that the quantization algorithms perform superiorly in terms of search accuracy, search time cost, and space cost. Finally, we introduce a few emerging topics.

Electrochemical Ammonia Synthesis via Nitrogen Reduction Reaction on a MoS2 Catalyst: Theoretical and Experimental Studies.

Abstract: The discovery of stable and noble-metal-free catalysts toward efficient electrochemical reduction of nitrogen (N2 ) to ammonia (NH3 ) is highly desired and significantly critical for the earth nitrogen cycle. Here, based on the theoretical predictions, MoS2 is first utilized to catalyze the N2 reduction reaction (NRR) under room temperature and atmospheric pressure. Electrochemical tests reveal that such catalyst achieves a high Faradaic efficiency (1.17%) and NH3 yield (8.08 × 10-11 mol s-1 cm-1 ) at -0.5 V versus reversible hydrogen electrode in 0.1 m Na2 SO4 . Even in acidic conditions, where strong hydrogen evolution reaction occurs, MoS2 is still active for the NRR. This work represents an important addition to the growing family of transition-metal-based catalysts with advanced performance in NRR.

STAMP: Short-Term Attention/Memory Priority Model for Session-based Recommendation

Abstract: Predicting users' actions based on anonymous sessions is a challenging problem in web-based behavioral modeling research, mainly due to the uncertainty of user behavior and the limited information. Recent advances in recurrent neural networks have led to promising approaches to solving this problem, with long short-term memory model proving effective in capturing users' general interests from previous clicks. However, none of the existing approaches explicitly take the effects of users' current actions on their next moves into account. In this study, we argue that a long-term memory model may be insufficient for modeling long sessions that usually contain user interests drift caused by unintended clicks. A novel short-term attention/memory priority model is proposed as a remedy, which is capable of capturing users' general interests from the long-term memory of a session context, whilst taking into account users' current interests from the short-term memory of the last-clicks. The validity and efficacy of the proposed attention mechanism is extensively evaluated on three benchmark data sets from the RecSys Challenge 2015 and CIKM Cup 2016. The numerical results show that our model achieves state-of-the-art performance in all the tests.

High-performance artificial nitrogen fixation at ambient conditions using a metal-free electrocatalyst

Abstract: Conversion of naturally abundant nitrogen to ammonia is a key (bio)chemical process to sustain life and represents a major challenge in chemistry and biology. Electrochemical reduction is emerging as a sustainable strategy for artificial nitrogen fixation at ambient conditions by tackling the hydrogen- and energy-intensive operations of the Haber–Bosch process. However, it is severely challenged by nitrogen activation and requires efficient catalysts for the nitrogen reduction reaction. Here we report that a boron carbide nanosheet acts as a metal-free catalyst for high-performance electrochemical nitrogen-to-ammonia fixation at ambient conditions. The catalyst can achieve a high ammonia yield of 26.57 μg h–1 mg–1cat. and a fairly high Faradaic efficiency of 15.95% at –0.75 V versus reversible hydrogen electrode, placing it among the most active aqueous-based nitrogen reduction reaction electrocatalysts. Notably, it also shows high electrochemical stability and excellent selectivity. The catalytic mechanism is assessed using density functional theory calculations. Electrochemical reduction of nitrogen is a promising route to industrial-scale nitrogen fixation at ambient conditions, but is challenged by activation of inert nitrogen. Here the authors report a metal-free catalyst that selectively reduces nitrogen to ammonia with high efficiency and stability.

Adaptive Fuzzy Neural Network Control for a Constrained Robot Using Impedance Learning

Abstract: This paper investigates adaptive fuzzy neural network (NN) control using impedance learning for a constrained robot, subject to unknown system dynamics, the effect of state constraints, and the uncertain compliant environment with which the robot comes into contact. A fuzzy NN learning algorithm is developed to identify the uncertain plant model. The prominent feature of the fuzzy NN is that there is no need to get the prior knowledge about the uncertainty and a sufficient amount of observed data. Also, impedance learning is introduced to tackle the interaction between the robot and its environment, so that the robot follows a desired destination generated by impedance learning. A barrier Lyapunov function is used to address the effect of state constraints. With the proposed control, the stability of the closed-loop system is achieved via Lyapunov’s stability theory, and the tracking performance is guaranteed under the condition of state constraints and uncertainty. Some simulation studies are carried out to illustrate the effectiveness of the proposed scheme.

Phosphorus-Doped Co3O4 Nanowire Array: A Highly Efficient Bifunctional Electrocatalyst for Overall Water Splitting

Abstract: It is vitally essential to design highly efficient and cost-effective bifunctional electrocatalysts toward water splitting. Herein, we report the development of P-doped Co3O4 nanowire array on nickel foam (P-Co3O4/NF) from Co3O4 nanowire array through low-temperature annealing, using NaH2PO2 as the P source. As a 3D catalyst, such P-Co3O4/NF demonstrates superior performance for oxygen evolution reaction with a low overpotential (260 mV at 20 mA cm–2), a small Tafel slope (60 mV dec–1), and a satisfying durability in 1.0 M KOH. Density functional theory calculations indicate that P-Co3O4 has a reaction free-energy value that is much smaller than that of pristine Co3O4 for the potential determining step of the oxygen evolution reaction. Such P-Co3O4/NF also performs efficiently for hydrogen evolution reaction, and a two-electrode alkaline electrolyzer assembled by P8.6-Co3O4/NF as both anode and cathode needs only 1.63 V to reach a water-splitting current of 10 mA cm–2.

Electrically conductive polymer composites for smart flexible strain sensors: a critical review

Abstract: The rapid development of wearable smart devices has contributed to the enormous demands for smart flexible strain sensors. However, to date, the poor stretchability and sensitivity of conventional metals or inorganic semiconductor-based strain sensors have restricted their application in this field to some extent, and hence many efforts have been devoted to find suitable candidates to overcome these limitations. Recently, novel resistive-type electrically conductive polymer composites (ECPCs)-based strain sensors have attracted attention based on their merits of light weight, flexibility, stretchability, and easy processing, thus showing great potential applications in the fields of human movement detection, artificial muscles, human–machine interfaces, soft robotic skin, etc. For ECPCs-based strain sensors, the conductive filler type and the phase morphology design have important influences on the sensing property. Meanwhile, to achieve a successful application toward wearable devices, several imperative features, including a self-healing capability, superhydrophobicity, and good light transmission, need to be considered. The aim of the present review is to critically review the progress of ECPCs-based strain sensors and to foresee their future development.

Boosted Electrocatalytic N2 Reduction to NH3 by Defect-Rich MoS2 Nanoflower

Abstract: The industrial artificial fixation of atmospheric N-2 to NH3 is carried out using the Haber-Bosch process that is not only energy-intensive but emits large amounts of greenhouse gas. Electrochemical reduction offers an environmentally benign and sustainable alternative for NH3 synthesis. Although Mo-dependent nitrogenases and molecular complexes effectively catalyze the N-2 fixation at ambient conditions, the development of a Mo-based nanocatalyst for highly performance electrochemical N-2 fixation still remains a key challenge. Here, greatly boosted electrocatalytic N-2 reduction to NH3 with excellent selectivity by defect-rich MoS2 nanoflowers is reported. In 0.1 m Na2SO4, this catalyst attains a high Faradic efficiency of 8.34% and a high NH3 yield of 29.28 mu g h(-1) mg(cat.)(-1) at (-)0.40 V versus reversible hydrogen electrode, much larger than those of defect-free counterpart (2.18% and 13.41 mu g h(-1) mg(cat.)(-1)), with strong electrochemical stability. Density functional theory calculations show that the potential determining step has a lower energy barrier (0.60 eV) for defect-rich catalyst than that of defect-free one (0.68 eV).

Vertical Co9S8 hollow nanowall arrays grown on a Celgard separator as a multifunctional polysulfide barrier for high-performance Li–S batteries

Abstract: Lithium–sulfur (Li–S) batteries have been regarded as one of the most promising next-generation energy-storage devices, due to their low cost and high theoretical energy density (2600 W h kg−1). However, the severe dissolution of lithium polysulfides (LiPSs) and the fatal shuttle effect of the sulfur cathode seriously hinder the practical applications of Li–S batteries. To address such issues, we present here, for the first time, a novel metal organic framework (MOF)-derived Co9S8 nanowall array with vertical hollow nanoarchitecture and high electrical conductivity, which is grown in situ on a Celgard separator (Co9S8–Celgard) via a feasible and scalable liquid-reaction approach, as an efficient barrier for LiPSs in Li–S batteries. Benefiting from the direct in situ growth of vertical Co9S8 hollow nanowall arrays as a multifunctional polar barrier, the Co9S8–Celgard separator possesses large surface area, excellent mechanical stability, and particularly strong LiPS-trapping ability via chemical and physical interactions. With these advantages, even with a pure sulfur cathode with a high sulfur loading of 5.6 mg cm−2, the Li–S cells with the Co9S8–Celgard separator exhibit outstanding electrochemical performance: the initial specific capacity is as high as 1385 mA h g−1 with a retention of 1190 mA h g−1 after 200 cycles. The cells deliver a high capacity of 530 mA h g−1 at a 1C rate (1675 mA g−1) even after an impressive number of 1000 cycles with an average capacity fade of only 0.039% per cycle, which is promising for long-term cycling application at high charge/discharge current densities, and pouch-type Li–S cells with the Co9S8–Celgard separator display excellent cycling performance. When the optimized cathode with the sulfur loading in well-designed yolk–shelled carbon@Fe3O4 (YSC@Fe3O4) nanoboxes is employed, the cell with Co9S8–Celgard delivers a high initial capacity of 986 mA h g−1 at a 1C rate with a capacity retention as high as 83.2% even after a remarkable number of 1500 cycles. This work presents a strategy to grow on the separator a multifunctional polar interlayer with unique nanoarchitecture and high conductivity to chemically and physically trap the LiPSs, thus significantly enhancing the performance of Li–S batteries.

TVQA: Localized, Compositional Video Question Answering

Abstract: Recent years have witnessed an increasing interest in image-based question-answering (QA) tasks. However, due to data limitations, there has been much less work on video-based QA. In this paper, we present TVQA, a large-scale video QA dataset based on 6 popular TV shows. TVQA consists of 152,545 QA pairs from 21,793 clips, spanning over 460 hours of video. Questions are designed to be compositional in nature, requiring systems to jointly localize relevant moments within a clip, comprehend subtitle-based dialogue, and recognize relevant visual concepts. We provide analyses of this new dataset as well as several baselines and a multi-stream end-to-end trainable neural network framework for the TVQA task. The dataset is publicly available at http://tvqa.cs.unc.edu.

Independently Recurrent Neural Network (IndRNN): Building A Longer and Deeper RNN

Abstract: Recurrent neural networks (RNNs) have been widely used for processing sequential data. However, RNNs are commonly difficult to train due to the well-known gradient vanishing and exploding problems and hard to learn long-term patterns. Long short-term memory (LSTM) and gated recurrent unit (GRU) were developed to address these problems, but the use of hyperbolic tangent and the sigmoid action functions results in gradient decay over layers. Consequently, construction of an efficiently trainable deep network is challenging. In addition, all the neurons in an RNN layer are entangled together and their behaviour is hard to interpret. To address these problems, a new type of RNN, referred to as independently recurrent neural network (IndRNN), is proposed in this paper, where neurons in the same layer are independent of each other and they are connected across layers. We have shown that an IndRNN can be easily regulated to prevent the gradient exploding and vanishing problems while allowing the network to learn long-term dependencies. Moreover, an IndRNN can work with non-saturated activation functions such as relu (rectified linear unit) and be still trained robustly. Multiple IndRNNs can be stacked to construct a network that is deeper than the existing RNNs. Experimental results have shown that the proposed IndRNN is able to process very long sequences (over 5000 time steps), can be used to construct very deep networks (21 layers used in the experiment) and still be trained robustly. Better performances have been achieved on various tasks by using IndRNNs compared with the traditional RNN and LSTM.

NAIS: Neural Attentive Item Similarity Model for Recommendation

Abstract: Item-to-item collaborative filtering ( aka. item-based CF) has been long used for building recommender systems in industrial settings, owing to its interpretability and efficiency in real-time personalization. It builds a user's profile as her historically interacted items, recommending new items that are similar to the user's profile. As such, the key to an item-based CF method is in the estimation of item similarities. Early approaches use statistical measures such as cosine similarity and Pearson coefficient to estimate item similarities, which are less accurate since they lack tailored optimization for the recommendation task. In recent years, several works attempt to learn item similarities from data, by expressing the similarity as an underlying model and estimating model parameters by optimizing a recommendation-aware objective function. While extensive efforts have been made to use shallow linear models for learning item similarities, there has been relatively less work exploring nonlinear neural network models for item-based CF. In this work, we propose a neural network model named Neural Attentive Item Similarity model (NAIS) for item-based CF. The key to our design of NAIS is an attention network, which is capable of distinguishing which historical items in a user profile are more important for a prediction. Compared to the state-of-the-art item-based CF method Factored Item Similarity Model (FISM) [1] , our NAIS has stronger representation power with only a few additional parameters brought by the attention network. Extensive experiments on two public benchmarks demonstrate the effectiveness of NAIS. This work is the first attempt that designs neural network models for item-based CF, opening up new research possibilities for future developments of neural recommender systems.

Materials and Structures toward Soft Electronics.

Abstract: Soft electronics are intensively studied as the integration of electronics with dynamic nonplanar surfaces has become necessary. Here, a discussion of the strategies in materials innovation and structural design to build soft electronic devices and systems is provided. For each strategy, the presentation focuses on the fundamental materials science and mechanics, and example device applications are highlighted where possible. Finally, perspectives on the key challenges and future directions of this field are presented.

iFeature: a Python package and web server for features extraction and selection from protein and peptide sequences.

Abstract: Summary Structural and physiochemical descriptors extracted from sequence data have been widely used to represent sequences and predict structural, functional, expression and interaction profiles of proteins and peptides as well as DNAs/RNAs. Here, we present iFeature, a versatile Python-based toolkit for generating various numerical feature representation schemes for both protein and peptide sequences. iFeature is capable of calculating and extracting a comprehensive spectrum of 18 major sequence encoding schemes that encompass 53 different types of feature descriptors. It also allows users to extract specific amino acid properties from the AAindex database. Furthermore, iFeature integrates 12 different types of commonly used feature clustering, selection and dimensionality reduction algorithms, greatly facilitating training, analysis and benchmarking of machine-learning models. The functionality of iFeature is made freely available via an online web server and a stand-alone toolkit. Availability and implementation http://iFeature.erc.monash.edu/; https://github.com/Superzchen/iFeature/. Supplementary information Supplementary data are available at Bioinformatics online.

Necklace-like Multishelled Hollow Spinel Oxides with Oxygen Vacancies for Efficient Water Electrolysis.

Abstract: The durability and reactivity of catalysts can be effectively and precisely controlled through the careful design and engineering of their surface structures and morphologies. Herein, we develop a novel “adsorption–calcination–reduction” strategy to synthesize spinel transitional metal oxides with a unique necklace-like multishelled hollow structure exploiting sacrificial templates of carbonaceous microspheres, including NiCo2O4 (NCO), CoMn2O4, and NiMn2O4. Importantly, benefiting from the unique structures and reduction treatment to offer rich oxygen vacancies, the unique reduced NCO (R-NCO) as a bifunctional electrocatalyst exhibits the dual characteristics of good stability as well as high electrocatalytic activity for both the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER). At 1.61 V cell voltage, a 10 mA cm–2 water splitting current density is obtained from the dual-electrode, alkaline water electrolyzer. Calculations based on density functional theory (DFT) reveal a mechanism ...

Unsupervised Deep Hashing with Similarity-Adaptive and Discrete Optimization

Abstract: Recent vision and learning studies show that learning compact hash codes can facilitate massive data processing with significantly reduced storage and computation. Particularly, learning deep hash functions has greatly improved the retrieval performance, typically under the semantic supervision. In contrast, current unsupervised deep hashing algorithms can hardly achieve satisfactory performance due to either the relaxed optimization or absence of similarity-sensitive objective. In this work, we propose a simple yet effective unsupervised hashing framework, named Similarity-Adaptive Deep Hashing (SADH), which alternatingly proceeds over three training modules: deep hash model training, similarity graph updating and binary code optimization. The key difference from the widely-used two-step hashing method is that the output representations of the learned deep model help update the similarity graph matrix, which is then used to improve the subsequent code optimization. In addition, for producing high-quality binary codes, we devise an effective discrete optimization algorithm which can directly handle the binary constraints with a general hashing loss. Extensive experiments validate the efficacy of SADH, which consistently outperforms the state-of-the-arts by large gaps.

A Hybrid Intelligent System Framework for the Prediction of Heart Disease Using Machine Learning Algorithms

Abstract: Heart disease is one of the most critical human diseases in the world and affects human life very badly. In heart disease, the heart is unable to push the required amount of blood to other parts of the body. Accurate and on time diagnosis of heart disease is important for heart failure prevention and treatment. The diagnosis of heart disease through traditional medical history has been considered as not reliable in many aspects. To classify the healthy people and people with heart disease, noninvasive-based methods such as machine learning are reliable and efficient. In the proposed study, we developed a machine-learning-based diagnosis system for heart disease prediction by using heart disease dataset. We used seven popular machine learning algorithms, three feature selection algorithms, the cross-validation method, and seven classifiers performance evaluation metrics such as classification accuracy, specificity, sensitivity, Matthews’ correlation coefficient, and execution time. The proposed system can easily identify and classify people with heart disease from healthy people. Additionally, receiver optimistic curves and area under the curves for each classifier was computed. We have discussed all of the classifiers, feature selection algorithms, preprocessing methods, validation method, and classifiers performance evaluation metrics used in this paper. The performance of the proposed system has been validated on full features and on a reduced set of features. The features reduction has an impact on classifiers performance in terms of accuracy and execution time of classifiers. The proposed machine-learning-based decision support system will assist the doctors to diagnosis heart patients efficiently.

Applications of Deep Reinforcement Learning in Communications and Networking: A Survey

Abstract: This paper presents a comprehensive literature review on applications of deep reinforcement learning in communications and networking. Modern networks, e.g., Internet of Things (IoT) and Unmanned Aerial Vehicle (UAV) networks, become more decentralized and autonomous. In such networks, network entities need to make decisions locally to maximize the network performance under uncertainty of network environment. Reinforcement learning has been efficiently used to enable the network entities to obtain the optimal policy including, e.g., decisions or actions, given their states when the state and action spaces are small. However, in complex and large-scale networks, the state and action spaces are usually large, and the reinforcement learning may not be able to find the optimal policy in reasonable time. Therefore, deep reinforcement learning, a combination of reinforcement learning with deep learning, has been developed to overcome the shortcomings. In this survey, we first give a tutorial of deep reinforcement learning from fundamental concepts to advanced models. Then, we review deep reinforcement learning approaches proposed to address emerging issues in communications and networking. The issues include dynamic network access, data rate control, wireless caching, data offloading, network security, and connectivity preservation which are all important to next generation networks such as 5G and beyond. Furthermore, we present applications of deep reinforcement learning for traffic routing, resource sharing, and data collection. Finally, we highlight important challenges, open issues, and future research directions of applying deep reinforcement learning.

A Survey on the Security of Blockchain Systems

Abstract: Since its inception, the blockchain technology has shown promising application prospects. From the initial cryptocurrency to the current smart contract, blockchain has been applied to many fields. Although there are some studies on the security and privacy issues of blockchain, there lacks a systematic examination on the security of blockchain systems. In this paper, we conduct a systematic study on the security threats to blockchain and survey the corresponding real attacks by examining popular blockchain systems. We also review the security enhancement solutions for blockchain, which could be used in the development of various blockchain systems, and suggest some future directions to stir research efforts into this area.

Modification strategies of TiO2 for potential applications in photocatalysis: a critical review

Abstract: TiO2 has received tremendous attention owing to its potential applications in the field of photocatalysis for solar fuel production and environmental remediation. This review mainly describes vario...

Additive Margin Softmax for Face Verification.

Abstract: In this letter, we propose a conceptually simple and intuitive learning objective function, i.e., additive margin softmax, for face verification. In general, face verification tasks can be viewed as metric learning problems, even though lots of face verification models are trained in classification schemes. It is possible when a large-margin strategy is introduced into the classification model to encourage intraclass variance minimization. As one alternative, angular softmax has been proposed to incorporate the margin. In this letter, we introduce another kind of margin to the softmax loss function, which is more intuitive and interpretable. Experiments on LFW and MegaFace show that our algorithm performs better when the evaluation criteria are designed for very low false alarm rate.

Perovskite Solar Cells with ZnO Electron-Transporting Materials

Abstract: Perovskite solar cells (PSCs) have developed rapidly over the past few years, and the power conversion efficiency of PSCs has exceeded 20%. Such high performance can be attributed to the unique properties of perovskite materials, such as high absorption over the visible range and long diffusion length. Due to the different diffusion lengths of holes and electrons, electron transporting materials (ETMs) used in PSCs play a critical role in PSCs performance. As an alternative to TiO2 ETM, ZnO materials have similar physical properties to TiO2 but with much higher electron mobility. In addition, there are many simple and facile methods to fabricate ZnO nanomaterials with low cost and energy consumption. This review focuses on recent developments in the use of ZnO ETM for PSCs. The fabrication methods of ZnO materials are briefly introduced. The influence of different ZnO ETMs on performance of PSCs is then reviewed. The limitations of ZnO ETM-based PSCs and some solutions to these challenges are also discussed. The review provides a systematic and comprehensive understanding of the influence of different ZnO ETMs on PSCs performance and potentially motivates further development of PSCs by extending the knowledge of ZnO-based PSCs to TiO2 -based PSCs.

Engineering graphene and TMDs based van der Waals heterostructures for photovoltaic and photoelectrochemical solar energy conversion

Abstract: Graphene and two-dimensional (2D) transition metal dichalcogenides (TMDs) have attracted significant interest due to their unique properties that cannot be obtained in their bulk counterparts. These atomically thin 2D materials have demonstrated strong light–matter interactions, tunable optical bandgap structures and unique structural and electrical properties, rendering possible the high conversion efficiency of solar energy with a minimal amount of active absorber material. The isolated 2D monolayer can be stacked into arbitrary van der Waals (vdWs) heterostructures without the need to consider lattice matching. Several combinations of 2D/3D and 2D/2D materials have been assembled to create vdWs heterojunctions for photovoltaic (PV) and photoelectrochemical (PEC) energy conversion. However, the complex, less-constrained, and more environmentally vulnerable interface in a vdWs heterojunction is different from that of a conventional, epitaxially grown heterojunction, engendering new challenges for surface and interface engineering. In this review, the physics of band alignment, the chemistry of surface modification and the behavior of photoexcited charge transfer at the interface during PV and PEC processes will be discussed. We will present a survey of the recent progress and challenges of 2D/3D and 2D/2D vdWs heterojunctions, with emphasis on their applicability to PV and PEC devices. Finally, we will discuss emerging issues yet to be explored for 2D materials to achieve high solar energy conversion efficiency and possible strategies to improve their performance.