Knowledge-based visual question answering(KB-VQA),requiring external world knowledge beyond the image for reasoning,is more challenging than traditional visual question answering.Recent works have demonstrated the eff...Knowledge-based visual question answering(KB-VQA),requiring external world knowledge beyond the image for reasoning,is more challenging than traditional visual question answering.Recent works have demonstrated the effectiveness of using a large(vision)language model as an implicit knowledge source to acquire the necessary information.However,the knowledge stored in large models(LMs)is often coarse-grained and inaccurate,causing questions requiring finer-grained information to be answered incorrectly.In this work,we propose a variational expectation-maximization(EM)framework that bootstraps the VQA performance of LMs with its own answer.In contrast to former VQA pipelines,we treat the outside knowledge as a latent variable.In the E-step,we approximate the posterior with two components:First,a rough answer,e.g.,a general description of the image,which is usually the strength of LMs,and second,a multi-modal neural retriever to retrieve question-specific knowledge from an external knowledge base.In the M-step,the training objective optimizes the ability of the original LMs to generate rough answers as well as refined answers based on the retrieved information.Extensive experiments show that our proposed framework,BootLM,has a strong retrieval ability and achieves state-of-the-art performance on knowledge-based VQA tasks.展开更多
Flexible wearable sensors with excellent electric response and self-powered capability have become an appealing hotspot for personal healthcare and human-machine interfaces.Here,based on triboelectric nanogenerator(TE...Flexible wearable sensors with excellent electric response and self-powered capability have become an appealing hotspot for personal healthcare and human-machine interfaces.Here,based on triboelectric nanogenerator(TENG),a flexible self-powered tactile sensor composed of micro-frustum-arrays-structured polydimethylsiloxane(PDMS)film/copper(Cu)electrodes,and poly(vinylidenefluoride-trifluoroethylene)(P(VDF-TrFE))nanofibers has been demonstrated.The TENG-based self-powered tactile sensor can generate electrical signals through the contact-separation process of two triboelectric layers under external mechanical stimuli.Due to the uniform and controllable micro-frustum-arrays structure fabricated by micro-electro-mechanical system(MEMS)process and the P(VDF-TrFE)nanofibers fabricated by electrostatic spinning,the flexible PDMS-based sensor presents high sensitivity of 2.97 V kPa^-1,stability of 40,000 cycles(no significant decay),response time of 60 ms at 1 Hz,low detection pressure of a water drop(~4 Pa,35 mg)and good linearity of 0.99231 in low pressure region.Since the PDMS film presents ultra-flexibility and excellent-biocompatibility,the sensor can be comfortably attached on human body.Furthermore,the tactile sensor can recognize various types of human body movements by the corresponding electrical signals.Therefore,the as-prepared TENGs are potential on the prospects of gesture detection,health assessment,human-machine interfaces and so on.展开更多
A peptide nucleic acid （PNA）-peptide conjugated molecule, T＇3（AKAE）2, was designed to have both a PNA segment for oligo- nucleotide binding and an ionic self-complementary peptide sequence for self-association. T...A peptide nucleic acid （PNA）-peptide conjugated molecule, T＇3（AKAE）2, was designed to have both a PNA segment for oligo- nucleotide binding and an ionic self-complementary peptide sequence for self-association. T＇3（AKAE）2 could co-assemble with oligoadenines （d（A）x） to form virus-like supramolecular structures whose morphology showed dependence on the chain length and rigidity of the d（A）x molecules. Smaller nanospheres with diameters of 13.0±2.0 nm were produced in the case of d（A）6. Wormlike aggregates with lengths of 20-50 nm and diameters of 15.0±2.5 nm were found in the cases of d（A）12, d（A）ls, d（A）24 and d（A）30. And larger spherical aggregates with diameters of 18±5 nm came into presence in the cases of d（A）36 and d（A）42＋. These nanostructures were suggested to be formed under a cooperative effect of base pair recognition and peptidic association. The study provides insights into the programmed assembly of a multi-components system as well as control of the size and shade of the co-assembled structures, which is of great significance in develouing gene/drug deliverv systems.展开更多
文摘Knowledge-based visual question answering(KB-VQA),requiring external world knowledge beyond the image for reasoning,is more challenging than traditional visual question answering.Recent works have demonstrated the effectiveness of using a large(vision)language model as an implicit knowledge source to acquire the necessary information.However,the knowledge stored in large models(LMs)is often coarse-grained and inaccurate,causing questions requiring finer-grained information to be answered incorrectly.In this work,we propose a variational expectation-maximization(EM)framework that bootstraps the VQA performance of LMs with its own answer.In contrast to former VQA pipelines,we treat the outside knowledge as a latent variable.In the E-step,we approximate the posterior with two components:First,a rough answer,e.g.,a general description of the image,which is usually the strength of LMs,and second,a multi-modal neural retriever to retrieve question-specific knowledge from an external knowledge base.In the M-step,the training objective optimizes the ability of the original LMs to generate rough answers as well as refined answers based on the retrieved information.Extensive experiments show that our proposed framework,BootLM,has a strong retrieval ability and achieves state-of-the-art performance on knowledge-based VQA tasks.
基金financially supported by the National Natural Science Foundation of China(51605449,51675493 and51705476)the National Key R&D Program of China(2018YFF0300605)+2 种基金Shanxi “1331 Project” Key Subject Construction(1331KSC)the Applied Fundamental Research Program of Shanxi Province(201601D021070)Zhangjiakou Science and Technology Research and Development Plan of Zhangjiakou City(1811009B-10)
文摘Flexible wearable sensors with excellent electric response and self-powered capability have become an appealing hotspot for personal healthcare and human-machine interfaces.Here,based on triboelectric nanogenerator(TENG),a flexible self-powered tactile sensor composed of micro-frustum-arrays-structured polydimethylsiloxane(PDMS)film/copper(Cu)electrodes,and poly(vinylidenefluoride-trifluoroethylene)(P(VDF-TrFE))nanofibers has been demonstrated.The TENG-based self-powered tactile sensor can generate electrical signals through the contact-separation process of two triboelectric layers under external mechanical stimuli.Due to the uniform and controllable micro-frustum-arrays structure fabricated by micro-electro-mechanical system(MEMS)process and the P(VDF-TrFE)nanofibers fabricated by electrostatic spinning,the flexible PDMS-based sensor presents high sensitivity of 2.97 V kPa^-1,stability of 40,000 cycles(no significant decay),response time of 60 ms at 1 Hz,low detection pressure of a water drop(~4 Pa,35 mg)and good linearity of 0.99231 in low pressure region.Since the PDMS film presents ultra-flexibility and excellent-biocompatibility,the sensor can be comfortably attached on human body.Furthermore,the tactile sensor can recognize various types of human body movements by the corresponding electrical signals.Therefore,the as-prepared TENGs are potential on the prospects of gesture detection,health assessment,human-machine interfaces and so on.
基金the National Natural Science Foundation of China (21473255, 21003160)the Fundamental Research Funds for the Central Universities (14CX05040A, 15CX05017A)
文摘A peptide nucleic acid （PNA）-peptide conjugated molecule, T＇3（AKAE）2, was designed to have both a PNA segment for oligo- nucleotide binding and an ionic self-complementary peptide sequence for self-association. T＇3（AKAE）2 could co-assemble with oligoadenines （d（A）x） to form virus-like supramolecular structures whose morphology showed dependence on the chain length and rigidity of the d（A）x molecules. Smaller nanospheres with diameters of 13.0±2.0 nm were produced in the case of d（A）6. Wormlike aggregates with lengths of 20-50 nm and diameters of 15.0±2.5 nm were found in the cases of d（A）12, d（A）ls, d（A）24 and d（A）30. And larger spherical aggregates with diameters of 18±5 nm came into presence in the cases of d（A）36 and d（A）42＋. These nanostructures were suggested to be formed under a cooperative effect of base pair recognition and peptidic association. The study provides insights into the programmed assembly of a multi-components system as well as control of the size and shade of the co-assembled structures, which is of great significance in develouing gene/drug deliverv systems.
