Search

Few-shot learning is an emerging yet challenging problem in which the goal is to achieve generalization from only few examples. Meta-learning tackles few-shot learning via the learning of prior knowledge shared across tasks and using it to learn new tasks. One of the most representative meta-learning algorithms is the model-agnostic meta-learning (MAML), which formulates prior knowledge as a common initialization, a shared starting point from where a learner can quickly adapt to unseen tasks. However, forcibly sharing an initialization can lead to conflicts among tasks and the compromised (undesired by tasks) location on optimization landscape, thereby hindering task adaptation. Furthermore, the degree of conflict is observed to vary not only among the tasks but also among the layers of a neural network. Thus, we propose task-and-layer-wise attenuation on the compromised initialization to reduce its adverse influence on task adaptation. As attenuation dynamically controls (or selectively forgets) the influence of the compromised prior knowledge for a given task and each layer, we name our method Learn to Forget (L2F). Experimental results demonstrate that the proposed method greatly improves the performance of the state-of-the-art MAML-based frameworks across diverse domains: few-shot classification, cross-domain few-shot classification, regression, reinforcement learning, and visual tracking.

Electrochemical oxygen reduction reactions (ORRs) require efficient, durable, and cost-effective electrocatalysts. In this regard, the development of metal-free catalysts with excellent activity is essential. Herein we report the development of a novel two-step heating route to produce metal-free N-doped carbon blacks. The prepared materials exhibit superior electrocatalytic performance for ORRs in a basic media, which is comparable to commercial Pt/C catalysts. Among the various heating temperatures tested, sequential two-step heating at 750 and 1000 °C produced the most efficient catalysts with onset and half-wave potentials of 0.97 and 0.82 V, respectively. The prepared materials show excellent selectivity for 4-electron transfer, durability, and stability in response to methanol poisoning. Control experiments with one-step heating and no N-doping confirm the superiority of the two-step heating process for producing efficient electrocatalysts.

The development of efficient catalysts for hydrogen evolution reaction (HER) presents a huge technical challenge. Graphitic carbon nitride (g-C3N4) is a promising metal-free, low cost, environment-friendly photocatalyst for HER that is driven by visible light. In this work, the authors provide new insight into the photocatalytic natures of g-C3N4 materials and their dependences on grain size, porosity, chemical structure, and photophysical properties. Three different precursors (urea, melamine, and dicyandiamide) and two gas atmospheres (air or N2) are used to produce various g-C3N4 materials. The use of urea and air leads to the formation of small grain C3N4 networks and porous structures with large surface areas. HER catalytic activity is promoted by large surface areas and the presence of terminal amine groups, and generation of small-sized Pt nanoparticle co-catalysts with narrow size distribution on the surface of g-C3N4. For samples with similar surface areas, band gaps and lifetimes of photogenerated charge carriers critically determine photocatalytic activities. By examining combinations of the above-mentioned factors, urea driven g-C3N4 produced in a N2 atmosphere is found to exhibit the best photocatalytic activity (up to 130 μmol h−1 g−1).

Graphene-based materials are promising as metal-free, cost-effective electrocatalysts owing to their feasibility for mass production and excellent electrochemical properties. Recently, heteroatom co-doping of graphene networks has been identified as a promising route to produce efficient electrocatalysts for the oxygen reduction reaction (ORR). In this study, we have synthesized a P, N co-doped graphene-based material (N-P-rG-O) using a two-step solution process in which graphene oxide is sequentially reacted with phytic acid and ammonium hydroxide in water. The N-P-rG-O catalyst exhibits an onset potential for the ORR in basic medium (0.1 M KOH) of 0.89 V (vs. the reversible hydrogen electrode), following a four-electron pathway and showing excellent stability against methanol poisoning in durability tests.

Hybrid materials containing single atoms or molecule-based active species immobilized on nanomaterials have been suggested as new, efficient catalytic systems for various reactions. In this study, novel hybrid materials consisting of molecularly dispersed Ni-based species on a graphitic carbon nitride (g-C3N4) framework are prepared, and their excellent electrocatalytic performance for the oxygen evolution reaction (OER) is discussed. Extensive chemical and structural characterizations confirm that the Ni-based species are attached and well-dispersed on the C3N4 network without agglomeration. In addition, results obtained from electrochemical characterization suggest that Ni-containing molecular entities dispersed on the C3N4 network are active species for the OER.

Graphitic carbon nitride (g-C3N4) has gained great attention as a material of promise for artificial photosynthesis. In place of synthesis of traditional three-dimensional g-C3N4 via polymerization of melamine or melem, recent studies seek to establish an alternative synthetic approach for two-dimensional g-C3N4 using a smaller precursor such as urea. However, the effectiveness of such a synthetic approach and resultant polymeric forms of g-C3N4 in this approach are still largely unknown. In this study, we present that solid-state NMR (SSNMR) analysis for 13C- and 15N-labeled g-C3N4 prepared from urea offers an unparalleled structural view for the heterogeneous in-plane structure of g-C3N4 and most likely for its moieties. We revealed that urea was successfully assembled in melem oligomers, which include extended oligomers involving six or more melem subunits. SSNMR, transmission electron micrograph, and ab initio calculation data suggested that the melem oligomer units were further extended into graphene...

Hybridization of organometallic complexes with carbon-based materials has been shown to enhance the catalytic performance in the oxygen reduction reaction (ORR). As chemical interactions between the metal centers and ligands are critical factors determining their tunable catalytic nature, it is important to understand the correlation between the chemical structure and catalytic nature and the correlation between the chemical structure and catalytic performance of the hybrids. In this work, the hybrids are synthesized by the reaction of an organometallic complex, Co(II)(acac)2 (acac = acetylacetonate), with refluxed graphene oxide (Re-G-O) materials containing controlled amounts of oxygen atoms at room temperature. Experimental characterization of the hybrids reveals that Co(II)(acac)2 is coordinated to oxygen containing groups in Re-G-O, generating organometallic species, Co–O4–O. These hybrids showed better electrocatalytic activity for the ORR in alkaline media than Co-free Re-G-O materials. The results show that the Co–O4 species is catalytically active for the ORR. We find that the best Co–O4 catalyst for the ORR is produced by the use of reduced graphene oxide with a medium level of reduction.

N-doped graphene-based materials show excellent promise as metal-free electrocatalysts for the oxygen reduction reaction (ORR). Because the N-doped graphene-based materials contain significant amounts of N and O atoms, it is important to understand the correlation between the chemical environments of these atoms and the catalytic performance of the materials. In this work, we prepared ammonia-reduced graphene oxide (A-rG-O) using a solution process and a set of A-rG-O samples thermally treated at high temperatures (300–700 °C). The chemical structure and ORR performances of A-rG-O materials were investigated.

A novel metal-free composite (GN) composed of two types of carbon-based nanomaterials, graphite oxide (GO) and 2D oxidized carbon nitride (OCN) nanodots was produced. Chemical and morphological characterizations reveal that GN contains a main component of GO with well-dispersed 2D OCN nanodots. GN shows enhanced photocatalytic performance for degrading an organic pollutant, Rhodamine B, under visible light.

Graphitic carbon nitride (g-C3 N4 ), synthesised by pyrolysis of different precursors (dicyandiamide, melamine and urea) under varying reaction conditions (air and nitrogen gas) is subjected to electrochemical studies for the elucidation of the inherent catalytic efficiency of the pristine material. Contrary to popular belief, pristine g-C3 N4 shows negligible, if any, enhancement in its electrochemical behaviour in this comprehensive study. Voltammetric analysis reveals g-C3 N4 to display similar catalytic efficiency to the unmodified glassy carbon electrode surface on which the bulk material was deposited. This highlights the non-catalytic nature of the pristine material and challenges the feasibility of using g-C3 N4 as a heterogeneous catalyst to deliver numerous promised applications.

Graphene-based materials are considered to be promising anodes for achieving high energy and power densities in Li-ion batteries (LIBs). Despite intensive research on reduced graphene oxide (rG-O) on this aspect, few have addressed the relationship between LIB performance and the degree of reduction of graphene oxide. In this work, we study the performance of LIB anodes using rG-O materials with different oxygen levels. A series of rG-O samples is produced by refluxing an aqueous graphene oxide suspension for different durations. The C/O ratios of the materials are found to increase gradually from approximately 1 to 6, with no heteroatoms other than oxygen atoms. The rG-O sample with the lowest degree of reduction delivers the highest anodic capacity and good durability.

Hybrid systems in which molecule-based active species are combined with nanoscale materials may offer valuable routes to enhance catalytic performances for electrocatalytic reactions. The development of rationally designed, cost-effective, efficient catalysts for the oxygen reduction reaction (ORR) is a crucial challenge for applications in fuel cells and metal-air batteries. A new hybrid ORR catalyst has been synthesized through a well-defined reaction between Co-based organometallic molecules and N-doped multiwalled carbon nanotubes (MWCNTs) at room temperature. The hybrid ORR catalyst shows excellent catalytic performance with an onset potential of 0.95 V [vs. the reversible hydrogen electrode (RHE)], superior durability, and good methanol tolerance. Chemical and structural characterization after many reaction cycles reveals that the Co-based organometallic species maintained the original structure of cobalt(II) acetylacetonate with coordination to the heteroatoms of the MWCNTs. A thorough electrochemical investigation indicates that the major catalytically active site is Co-O4 -NCNT .

The authors studied the CO2 adsorption performances of borane modified graphene-based materials (B-rG-O). B-rG-O powder was fabricated by reacting graphene oxide with a borane–THF adduct and showed good CO2 adsorption capacity (1.82 mmol/g at 1 atm) and recyclability as determined using a CO2 isotherm and breakthrough experiments respectively. The roles of boron moieties for CO2 adsorption were studied using theoretical calculation and spectroscopic experiments.

The production of colloidal suspensions of graphene-based nano-platelets in large scale is highly important for their use in practical applications. In this work, we developed a new route for generation of colloidal suspensions of N-modified graphene-based nano-platelets (A-rG-O) with high concentration in water or organic solvent/water mixed systems by the reaction between graphene oxide and ammonium hydroxide. Graphene oxide was reduced by the ammonium hydroxide and N atoms (up to 6 at%) were introduced into A-rG-O materials using solution process without further thermal treatment. Such A-rG-O nano-platelets were well dispersed as individual layers in both systems. Macroscopic free-standing A-rG-O paper materials with good electrical conductivity were produced by filtration of such colloidal suspensions.

Graphitic carbon nitride (g-C3N4) and its derivatives are promising candidates as catalysts or supports for photocatalytic applications. Since they are typically produced by polymerization or condensation of monomers under high temperature and high pressure, development of a cost-effective, solution-based, low-temperature method of production is important. Herein, novel hybrid materials composed of g-C3N4 and reduced graphene oxide are produced using a simple reaction between graphene oxide and cyanamide using a solution-based process. During the reaction, reduction of graphene oxide and graphene oxide-assisted generation of g-C3N4 occurred simultaneously. These hybrids show good photocatalytic performance for the removal of organic dyes under one sun solar light illumination.

We developed a new route for scalable production of graphene-based nano-platelets containing extremely low amounts of heteroatoms via treatment of graphene oxide suspensions with ammonia borane at low temperatures without further thermal annealing. The C/O atomic ratio measured by combustion-based elemental analysis is the highest value among published data on graphene-based nano-platelets using solution processes. Resulting materials showed a very high electrical conductivity.

As the device size decreases, the reliability degradation caused by soft-errors and multiple component error due to a single soft-error are becoming serious problems in VLSIs. In this study, we propose a method to synthesize single soft-error tolerant datapaths based on soft-error detection by duplication and comparison and correction by retry. Under the assumption that a single soft-error does not affect beyond a certain spatial and time range, we consider adjacency constraint between components in datapath. The major contributions of this research are (i) adjacency constraint between components to employ appropriate soft-error tolerance and achieve more effective use of resources and (ii) an ILP based resource binding considering adjacency constraint to minimize the increment of the chip size due to introducing the new constraint.

Photocatalysts use sustainable solar light energy to trigger various catalytic reactions. Metal-free nanomaterials have been suggested as cost-effective and environmentally friendly photocatalysts. In this work, we propose thickness-controlled graphite oxide (GO) as a metal-free photocatalyst, which is produced by exfoliating thick GO particles via stirring and sonication. All GO samples exhibit photocatalytic activity for degrading an organic pollutant, rhodamine B under visible light, and the thickest sample shows the best catalytic performance. UV-vis-NIR diffuse reflectance absorption spectra indicate that thicker GO samples absorb more vis-NIR light than thinner ones. Density-functional theory calculations show that GO has a much smaller band gap than that of single-layer graphene oxide, and thus suggest that the largely-reduced band gap is responsible for this trend of light absorption.

Macroscopic, freestanding graphene-based paper-like materials are of interest for use as mechanically strong, stiff, and flexible and electrically conductive materials. Chemically reduced graphene oxide paper shows promise for such applications. In this work, we studied the mechanical and electrical properties of a set of paper materials prepared by filtration of homogeneous colloidal suspensions of hydrazine-reduced graphene oxide with different concentrations. Young’s modulus, fracture strength, and fracture strain of each type of sample was determined by tensile tests. The paper sample prepared from the colloidal suspension with the lowest concentration of reduced graphene oxide platelets had the highest modulus and fracture strength and showed the smoothest surface morphology. The electrical conductivity measured by the four-probe measurement method increased as the concentration was increased.

Chemically modified graphenes (CMGs) are promising candidates for a wide range of applications. Graphite oxide (GO) is most commonly used to produce CMGs. We note that residual acetone can produce dangerous explosives during the synthesis of GO. Addition of acetone produced acetone peroxides via a reaction with H 2 O 2 and H 2 SO 4 , which are used in the Hummers/modified Hummers methods of making GO. The use of Na 2 S 2 O 3 instead of H 2 O 2 yielded GO without making explosives.

Three-dimensional (3D) carbon nitride (C3 N4 )-based materials show excellent performance in a wide range of applications because of their suitable band structures. To realize the great promise of two-dimensional (2D) allotropes of various 3D materials, it is highly important to develop routes for the production of 2D C3 N4 materials, which are one-atom thick, in order to understand their intrinsic properties and identify their possible applications. In this work, water-dispersible, atomically thin, and small carbon nitride nanodots were produced using the chemical oxidation of graphitic C3 N4 . Various analyses, including X-ray diffraction, X-ray photoelectron, Fourier-transform infrared spectroscopy, and combustion-based elemental analysis, and thermogravimetric analysis, confirmed the production of 3D oxidized C3 N4 materials. The 2D C3 N4 nanodots were successfully exfoliated as individual single layers; their lateral dimension was several tens of nanometers. They showed strong photoluminescence in the visible region as well as excellent performances as cell-imaging probes in an in vitro study using confocal fluorescence microscopy.

Chemically modified graphene (CMG) nanoplatelets have shown great promise in various applications due to their electrical properties and high surface area. Chemical doping is one of the most effective methods to tune the electronic properties of graphene materials. In this work, novel B-doped nanoplatelets (borane-reduced graphene oxide, B-rG-O) were produced on a large scale via the reduction of graphene oxide by a borane-tetrahydrofuran adduct under reflux, and their use for supercapacitor electrodes was studied. This is the first report on the production of B-doped graphene nanoplatelets from a solution process and on the use of B-doped graphene materials in supercapacitors. The B-rG-O had a high specific surface area of 466 m(2)/g and showed excellent supercapacitor performance including a high specific capacitance of 200 F/g in aqueous electrolyte as well as superior surface area-normalized capacitance to typical carbon-based supercapacitor materials and good stability after 4500 cycles. Two- and three-electrode cell measurements showed that energy storage in the B-rG-O supercapacitors was contributed by ion adsorption on the surface of the nanoplatelets in addition to electrochemical redox reactions.

Catalytic natures of organometallic catalysts are modulated by coordinating organic ligands with proper steric and electronic properties to metal centers. Carbon-based nanomaterials such as graphene nanoplatelets are used with and without N-doping and multiwalled carbon nanotube as a ligand for ethylene polymerizations. Zirconocenes or titanocenes are immobilized on such nanomaterials. Polyethylenes (PEs) produced by such hybrids show a great increase in molecular weight relative to those produced by free catalysts. Specially, ultra-high-molecular-weight PEs are produced from the polymerizations at low temperature using the hybrid with N-doped graphene nanoplatelets. This result shows that such nanomaterials act a crucial role to tune the catalytic natures of metallocenes.

We present efficient simplified behavioral models for power amplifier(PA) and its digital predistorter(DPD). Even though the many promising models have been proposed, more rigorous derivation and improvements in terms of performance and complexity are needed. In this paper, new approach is attempted to tackle the above. After building basis set based on simple rules, the new efficient and simplified models are derived by adding only the appropriate basis into the previous composite bases. The experimental results show that the proposed models for the PA and DPD outperform the conventional models.

Development of new systems with high catalytic performances in the oxygen reduction reaction (ORR) for practical applications in fuel cells and metal–air batteries is a challenge. We develop a one-pot solution method for producing a novel hybrid material consisting of Ir species anchored on N-doped graphene. The hybrid is synthesized by reacting graphene oxide with IrCl3 xH2O in dimethylformamide under reflux. Chemical and structural analyses confirm the attachment of Ir atoms to the N and O atoms of the N-doped graphene-based materials. The hybrid shows a good electrocatalytic performance for the ORR in alkaline media, with an onset potential of 0.88 V (versus the reversible hydrogen electrode), high long-term durability, and good tolerance for methanol poisoning.