Model distillation aims to distill the knowledge of a complex model into a simpler one. In this paper, we consider an alternative formulation called dataset distillation: we keep the model fixed and instead attempt to distill the knowledge from a large training dataset into a small one. The idea is to synthesize a small number of data points that do not need to come from the correct data distribution, but will, when given to the learning algorithm as training data, approximate the model trained on the original data. For example, we show that it is possible to compress 60, 000 MNIST training images into just 10 synthetic distilled images (one per class) and achieve close to original performance with only a few gradient descent steps, given a fixed network initialization. We evaluate our method in various initialization settings and with different learning objectives. Experiments on multiple datasets show the advantage of our approach compared to alternative methods.
We present a machine learning algorithm that takes as input a 2D RGB image and synthesizes a 4D RGBD light field (color and depth of the scene in each ray direction). For training, we introduce the largest public light field dataset, consisting of over 3300 plenoptic camera light fields of scenes containing flowers and plants. Our synthesis pipeline consists of a convolutional neural network (CNN) that estimates scene geometry, a stage that renders a Lambertian light field using that geometry, and a second CNN that predicts occluded rays and non-Lambertian effects. Our algorithm builds on recent view synthesis methods, but is unique in predicting RGBD for each light field ray and improving unsupervised single image depth estimation by enforcing consistency of ray depths that should intersect the same scene point.
The electrochemical oxidation of small molecules to generate value‐added products has gained enormous interest in recent years because of the advantages of benign operation conditions, high conversion efficiency and selectivity, the absence of external oxidizing agents, and eco‐friendliness. Coupling the electrochemical oxidation of small molecules to replace oxygen evolution reaction (OER) at the anode and the hydrogen evolution reaction (HER) at the cathode in an electrolyzer would simultaneously realize the generation of high‐value chemicals or pollutant degradation and the highly efficient production of hydrogen. This Minireview presents an introduction on small‐molecule choice and design strategies of electrocatalysts as well as recent breakthroughs achieved in the highly efficient production of hydrogen. Finally, challenges and future orientations are highlighted.
Herein we firstly introduce a straightforward, scalable and technologically relevant strategy to manufacture charged porous polymer membranes (CPMs) in a controllable manner. The pore sizes and porous architectures of CPMs are well-controlled by rational choice of anions in poly(ionic liquid)s (PILs). Continuously, heteroatom-doped hierarchically porous carbon membrane (HCMs) can be readily fabricated via morphology-maintaining carbonization of asprepared CPMs. These HCMs being as photothermal membranes exhibited excellent performance for solar seawater desalination, representing a promising strategy to construct advanced functional nanomaterials for portable water production technologies.Charged porous polymer membranes (CPMs) have been attracting widespread attention in both academia and industry because they can serve as a multifunctional platform beyond merely filtration membranes. 1 Particularly, the synergy of pore confinement, charges and flexible design in surface chemistry endows them versatile for device fabrication, 2 separation, 3 controlled release, 4 catalyst supports, 5 bio-interfacing, 6 sensors, 7 etc. However, their unique charge nature of CPMs inherently challenges the state-of-the-art membrane fabrication techniques, 8 retarding their development and utilization. There are generally three strategies to fabricate CPMs: (i) selfassembly and dewetting of block copolymers or their blends; 9 (ii) electrostatic layer-by-layer assembly under carefully designed conditions on a laboratory scale 10 ; (iii) electrostatic complexation, especially between a hydrophobic polycation and a hydrophilic polyanion. 11 While the former two suffer from considerable time-and labor-demand and difficulties in obtaining freestanding, interconnected porous membranes, the latter is free of these problems but much
Electrochemically functional porous membranes of low cost are appealing in various electrochemical devices used in modern environmental and energy technologies.H erein we describe as calable strategy to construct electrochemically active,h ierarchically porous carbon membranes containing atomically dispersed semi-metallic Se,d enoted SeNCM. The isolated Se atoms were stabilized by carbon atoms in the form of ah exatomic ring structure,i nw hich the Se atoms were located at the edges of graphitic domains in SeNCM. This configuration is different from that of previously reported transition/noble metal single atom catalysts.T he positively charged Se,e nlarged graphitic layers,r obust electrochemical nature of SeNCM endow them with excellent catalytic activity that is superior to state-of-the-art commercial Pt/C catalyst. It also has long-term operational stability for hydrazine oxidation reaction in practical hydrazine fuel cell.Supportinginformation and the ORCID identification number(s) for the author(s) of this article can be found under: https://doi.
Main‐group (s‐ and p‐block) metals are generally regarded as catalytically inactive due to the delocalized s/p‐band. Herein, we successfully synthesized a p‐block antimony single‐atom catalyst (Sb SAC) with the Sb−N4 configuration for efficient catalysis of the oxygen reduction reaction (ORR). The obtained Sb SAC exhibits superior ORR activity with a half‐wave potential of 0.86 V and excellent stability, which outperforms most transition‐metal (TM, d‐block) based SACs and commercial Pt/C. In addition, it presents an excellent power density of 184.6 mW cm−2 and a high specific capacity (803.5 mAh g−1) in Zn–air battery. Both experiment and theoretical calculation manifest that the active catalytic sites are positively charged Sb−N4 single‐metal sites, which have closed d shells. Density of states (DOS) results unveil the p orbital of the atomically dispersed Sb cation in Sb SAC can easily interact with O2‐p orbital to form hybrid states, facilitating the charge transfer and generating appropriate adsorption strength for oxygen intermediates, lowering the energy barrier and modulating the rate‐determining step. This work sheds light on the atomic‐level preparing p‐block Sb metal catalyst for highly active ORR, and further provides valuable guidelines for the rational design of other main‐group‐metal SACs.
Contrastive representation learning has been outstandingly successful in practice. In this work, we identify two key properties related to the contrastive loss: (1) alignment (closeness) of features from positive pairs, and (2) uniformity of the induced distribution of the (normalized) features on the hypersphere. We prove that, asymptotically, the contrastive loss optimizes these properties, and analyze their positive effects on downstream tasks. Empirically, we introduce an optimizable metric to quantify each property. Extensive experiments on standard vision and language datasets confirm the strong agreement between both metrics and downstream task performance. Remarkably, directly optimizing for these two metrics leads to representations with comparable or better performance at downstream tasks than contrastive learning. Project
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
334 Leonard St
Brooklyn, NY 11211
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.