Modal Transfer Research Articles

Semi-supervised cross-modal representation learning with GAN-based Asymmetric Transfer Network

In this paper, we proposed a semi-supervised common representation learning method with GAN-based Asymmetric Transfer Network (GATN) for cross modality retrieval. GATN utilizes the asymmetric pipeline to guarantee the semantic consistency and adopt (Generative Adversarial Network) GAN to fit the distributions of different modalities. Specifically, the common representation learning across modalities includes two stages: (1) the first stage, GATN trains source mapping network to learn the semantic representation of text modality by supervised method; and (2) the second stage, GAN-based unsupervised modality transfer method is proposed to guide the training of target mapping network, which includes generative network (target mapping network) and discriminative network. Experimental results on three widely-used benchmarks show that GATN have achieved better performance comparing with several existing state-of-the-art methods.

Transfer of Dry-Land Resistance Training Modalities to Swimming Performance.

A great number of studies focusing on the effects of dry-land resistance training interventions on swimming performance remain inconclusive. It is suggested that transferability of dry-land strength gains to swimming performance appear when dry-land resistance training programs are swim-specific. The main aim of this study was to compare the effects of specific dry-land resistance training on an ergometer with traditional dry-land exercises, and to determine how much of the resistance training effects were transferred to specific swimming conditions. The study included a group of 26 youth competitive male swimmers (age 15.7 ± 0.5 years, height 174.6 ± 6.6 cm, weight 68.4 ± 8.2 kg, training experience 5.8±0.7 years) of regional level. They were randomly allocated to one of two groups: experimental (E) and control (T). Both groups were involved in a 12-week dry-land resistance training concentrated on increasing muscular strength and power output of the upper limbs. Group E used a specialized ergometer (JBA – Zbigniew Staniak), while group T performed traditional resistance exercises. The program consisted of 10 sets of 30 s of exercise with 30 s rest intervals between each set. A two-way repeated measures ANOVA with Tukey HSD post hoc comparisons was used to determine if any significant differences existed between training groups across pretest and posttest conditions. The significance level was set at p ≤ 0.05. Dry-land resistance training modalities were the only differences in training between both groups. Our findings show that rates of transfer are much higher in group E than in group T, which resulted in a significant increase in swimming velocity (by 4.32%, p<0.001; ES=1.23, and 2.78%, p<0.003, ES=0.31, respectively).

The link between bike sharing and subway use during the COVID-19 pandemic: The case-study of New York's Citi Bike

The full societal impact COVID-19 pandemic is laid bare in urban mobility patterns. This research explored the recently published data on the operation of subway and bike share systems (BSS) during the COVID-19 outbreak in New York city, providing evidence on its impact over urban transport systems, but also on how its different components can work in conjunction. The BSS has proved to be more resilient than the subway system, with a less significant ridership drop (71% vs 90% ridership drop and 50% decrease on the ridership ratio) and an increase on its trips' average duration (from 13 min to 19 min per trip). Moreover, the study found evidence of a modal transfer from some subway users to the bike sharing system. The first effects of the free BSS programs aimed at essential service workers were also evaluated. BSS can improve the resilience of urban transport systems to disruptive events. Overall, this paper offers clues on how bike sharing, and cycling in general, can support the transition to a post-coronavirus society.

Augmented Adversarial Training for Cross-Modal Retrieval

Cross-modal retrieval has received considerable attention in recent years. The core of cross-modal retrieval is to find a representation space to align data from different modalities according to their semantics. In this paper, we propose a cross-modal retrieval method that aligns data from different modalities by transferring one source modality to another target modality with augmented adversarial training. To preserve the semantic meaning in the modality transfer process, we employ the idea of conditional GANs and augment it. The key idea is to incorporate semantic information from the label space into the adversarial training process by sampling more semantic relevant and irrelevant source-target sample pairs. The augmented sample pairs improve the alignment from two aspects. First, relevant source-target sample pairs provide more training samples, leading to a better guidance of the alignment of fake targets and true paired targets. Second, relevant and irrelevant source-target sample pairs teach the discriminator to better distinguish true relevant pairs from fake relevant pairs, which guides the generator to better transfer from the source modality to the target modality. Extensive experiments compared with state-of-the-art methods show the promising power of our approach.

Solutions for Improving Transit through Intermodal Passenger Terminals

According to the transport White Paper “better modal choices will result from greater integration of the modal networks: airports, ports, railway, metro and bus stations, should increasingly be linked and transformed into multimodal connection platforms for passengers.”Transport terminals are the junction points of local, regional or inter-regional transport. The main objective of a transport terminal is to serve the passengers’ demand by maintaining their safety and comfort. Intermodal terminals have emerged as a response or solution to the demands of sustainable development (reducing congestion should be the main objective), Bucharest ranks 5th in the world at congestion (according to Tomtom Traffic Index). The use of intermodality in passenger transport significantly reduces congestion in cities.Intermodal passenger terminals lead to reduced congestion but impair passenger quality parameters such as time, comfort, convenience and even travel’s cost because each mode of transport has its own pricing principles.The problem of reducing the transit time through these intermodal terminals is the key in making the public transport more attractive; the way the terminals are designed and organized leads to their intensive use or not. The paper presents a model for designing intermodal passenger terminals that encourage modal transfer. The model’s framework suppose the correlation of the transport modes timetables, placing the platforms so that the movements between them are as small as possible, tariff integration and the use of an unique ticket, etc.The model presented within the paper has been tested on a hypothetical intermodal passenger terminal that has benefitted by transport modes schedules correlation through a linear programming with integer numbers model.

HMTL: Heterogeneous Modality Transfer Learning for Audio-Visual Sentiment Analysis

Multimodal sentiment analysis is an extended approach to traditional language-based sentiment analysis, which uses other relevant modality data. Multimodal sentiment analysis usually applies visual, textual, and acoustic representations for sentiment prediction. Recently, various data fusion methodologies have been proposed for multimodal sentiment analysis. In most cases, textual modality plays a major role, and visual and acoustic modalities are used as auxiliary sources for multimodal sentiment analysis. However, in general multimedia such as video, text transcripts of an individual's speech are not provided. Research on an audio-visual sentiment analysis methodology that does not depend on text modality is essential for multimodal sentiment analysis in real-world industrial applications. Therefore, it is important to improve audio-visual sentiment analysis because it currently exhibits lower performance than multimodal sentiment analysis, including text modality. In this paper, we propose heterogeneous modality transfer learning (HMTL) to utilize the knowledge of aligned text data as a source modality in transfer learning to improve audio-visual sentiment analysis performance. Our approach uses a decoder and adversarial learning techniques to reduce the gap between the source and target modalities in the embedded space for multimodal representation. Our proposed methodology experimentally outperformed recent unimodal and bimodal audio-visual sentiment analysis achievements.

ANALISIS KUALITAS PELAYANAN FASILITAS TERMINAL KAMPUNG RAMBUTAN BERDASARKAN TINGKAT KEPUASAN PENGGUNA

Keberadaan terminal dalam jaringan transportasi sangat penting, sebab merupakan tempat perpindahan moda. Terminal bus Kampung Rambutan merupakan terminal Tipe-A yang melayani hilir mudik penumpang dari dan ke berbagai kota di provinsi lain maupun berbagai tempat di dalam DKI Jakarta. Sehingga kinerja terminal adalah hal yang sangat penting agar dapat menunjang kelancaran, keamanan dan kenyamanan penumpang dalam berpindah moda. Penelitian ini bertujuan untuk mengetahui kualitas pelayanan terminal Kampung Rambutan terhadap kepuasan pengguna, mengetahui kinerja eksisting terminal berdasarkan Standar Pelayanan Minimal (SPM), dan memberi rekomendasi perbaikan pelayanan fasilitas Terminal Kampung Rambutan. Survey dilakukan dengan penyebaran kuisioner kepada pengguna terminal (penumpang) dan inventarisasi kondisi eksisting terminal. Analisis dilakukan dengan menggunakan Importance-Performance Analysis (IPA) dan statistik deskriptif. Hasil analisis menunjukkan bahwa terdapat beberapa indikator menjadi prioritas utama yang harus diperbaiki menurut pengguna terminal yaitu fasilitas bagi pengguna berkebutuhan khusus (difabel), penyampaian informasi keberangkatan dan kedatangan bus, keamanan dan kenyamanan.

DUAL-GLOW: Conditional Flow-Based Generative Model for Modality Transfer.

Positron emission tomography (PET) imaging is an imaging modality for diagnosing a number of neurological diseases. In contrast to Magnetic Resonance Imaging (MRI), PET is costly and involves injecting a radioactive substance into the patient. Motivated by developments in modality transfer in vision, we study the generation of certain types of PET images from MRI data. We derive new flow-based generative models which we show perform well in this small sample size regime (much smaller than dataset sizes available in standard vision tasks). Our formulation, DUAL-GLOW, is based on two invertible networks and a relation network that maps the latent spaces to each other. We discuss how given the prior distribution, learning the conditional distribution of PET given the MRI image reduces to obtaining the conditional distribution between the two latent codes w.r.t. the two image types. We also extend our framework to leverage "side" information (or attributes) when available. By controlling the PET generation through "conditioning" on age, our model is also able to capture brain FDG-PET (hypometabolism) changes, as a function of age. We present experiments on the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset with 826 subjects, and obtain good performance in PET image synthesis, qualitatively and quantitatively better than recent works.

Improving the malignancy characterization of hepatocellular carcinoma using deeply supervised cross modal transfer learning for non-enhanced MR.

The malignancy characterization of hepatocellular carcinoma (HCC) is remarkably significant in clinical practice. In this work, we propose a deeply supervised cross modal transfer learning method to remarkably improve the malignancy characterization of HCC based on non-enhanced MR. First, we use samples of non-enhanced and contrast-enhanced MR for pre-training a deep learning network to learn the cross modal relationship between the non-enhanced modal and enhanced modal. Then, the parameters of the pre-trained across modal representation are transferred to a second deep learning model for fine-tuning based only on non-enhanced MR for malignancy characterization of HCC. Specifically, a deeply supervised network is designed to enhance the stability of the second deep learning model and further improve the performance of lesion characterization. Importantly, only non-enhanced MR of HCC is required for the malignancy characterization in the training and test phase of the second deep learning model. Experiments of one hundred and fifteen clinical HCCs demonstrate that the proposed deeply supervised cross modal transfer learning method can significantly improve the malignancy characterization of HCC based on non-enhanced MR.

PROBABILITY MODAL TRANSFER TO THE FORWARDER AT TERMINAL LAMONG BAY SURABAYA

Currently the mobilization of freight transport from Surabaya to Jakarta or otherwise served by three types of modes transport, including truck, train, and marine vessels. But the majority of the freight transport cargo in the corridor Surabaya – Jakarta is served by trucking so that the impact on the capacity of roads and road load increases. The purpose of this study was to estimate the probability EMKL service manager / Forwarder to switch modes from container truck to railway mode container. The methods used to analyze the research data is a binary logistic regression. The results of these studies indicate that the travel time during the 3-day trip by truck containers will potentially probability modal transfer containers from trucks to rail containers reached 87.49%, and the more need for the use of container sizes of 20 ft. at a cost of Rp. 2,500,000 will potentially modal transfer containers from trucks to rail container by 82.34%, more than the container size of 40 ft. at a cost of Rp. 4,000,000 by 50%.

Broadband and low-loss mode scramblers using CO2-laser inscribed long-period gratings.

We demonstrate broadband and low-loss three-mode and six-mode scramblers employing CO2-laser inscribed long-period gratings (LPGs) for space-division multiplexing. Step-index (SI) few-mode fibers are used to avoid mode coupling to the cladding modes. We characterize the mode scramblers using a swept-wavelength interferometer. Mode-dependent loss (MDL) and modal transfer matrices over the C+L band are presented. Demonstrated LPGs with negligible MDL and low insertion loss contributed to high-performance CO2-laser inscription. The total MDLs induced by the SI fiber with LPGs in three-mode and six-mode scramblers are measured to be 2 and 4dB, respectively.

Development of planar rigid-elastic coupling tyre model with analytical multi-stiffness sidewall

In this study, combining the membrane feature with inflation pressure and the structural deformation caused by sidewall curvature, rigid-elastic coupling tyre model with analytical multi-stiffness sidewall is proposed for a heavy-loaded radial tyre with a large section ratio. The membrane pre-tension of sidewall arc resulting from inflation pressure is investigated. By means of virtual work principle, the structural deformation of sidewall curved arc resulting from the arc curvature, including stretching, bending and shearing deformation is derived. The structural stiffness caused by the sidewall curvature and membrane pre-tension caused by the inflation pressure are combined for the multi-stiffness sidewall model. The influence of the sidewall structural and geometrical parameters on the sidewall multi-stiffness, modal resonant frequency and transfer function is researched and discussed. The non-linear characteristic of sidewall multi-stiffness with respect to the sidewall radial deformation is investigated. Experimental and theoretical results show that: (1) the multi-stiffness of sidewall can characterise the membrane-tension stiffness caused by inflation pressure and the structural stiffness led by the sidewall curvature and material properties; (2) the different multi-stiffnesses of upper and lower sidewall arcs results from the different interval angles; (3) the multi-stiffness of sidewall is non-linear to the radial sidewall deformation. Taking the flexible deformation of tyre carcass and the analytical multi-stiffness of tyre sidewall into consideration, rigid-elastic coupling tyre model with multi-stiffness sidewall is suitable for the heavy-loaded radial tyre with a large section ratio or tyres under impulsive loading and large deformation.

Fast mean and variance computation of the diffuse sound transmission through finite-sized thick and layered wall and floor systems

A method is developed for computing the mean and variance of the diffuse field sound transmission loss of finite-sized layered wall and floor systems that consist of solid, fluid and/or poroelastic layers. This is achieved by coupling a transfer matrix model of the wall or floor to statistical energy analysis subsystem models of the adjacent room volumes. The modal behavior of the wall is approximately accounted for by projecting the wall displacement onto a set of sinusoidal lateral basis functions. This hybrid modal transfer matrix-statistical energy analysis method is validated on multiple wall systems: a thin steel plate, a polymethyl methacrylate panel, a thick brick wall, a sandwich panel, a double-leaf wall with poro-elastic material in the cavity, and a double glazing. The predictions are compared with experimental data and with results obtained using alternative prediction methods such as the transfer matrix method with spatial windowing, the hybrid wave based-transfer matrix method, and the hybrid finite element-statistical energy analysis method. These comparisons confirm the prediction accuracy of the proposed method and the computational efficiency against the conventional hybrid finite element-statistical energy analysis method.

Stem cell tracking: toward clinical application in oncology?

Noninvasive cellular imaging allows the tracking of grafted cells as well as the monitoring of their migration, suggesting potential applications to track both cancer and therapeutic stem cells. Cell tracking can be performed by two approaches: direct labeling (cells are labeled with tags) and indirect labeling (cells are transfected with a reporter gene and visualized after administration of a reporter probe). Techniques for in vivo detection of grafted cells include optic imaging, nuclear medicine imaging, magnetic resonance imaging, microCT imaging and ultrasound imaging. The ideal imaging modality would bring together high sensitivity, high resolution and low toxicity. All of the available imaging methods are based on different principles, have different properties and different limitations, so several of them can be considered complementary. Transfer of these preclinical cellular imaging modalities to stem cells has already been reported, and transfer to clinical practice within the next years can be reasonably considered.

Decentralized semi-active damping of free structural vibrations by means of structural nodes with an on/off ability to transmit moments

Abstract This paper proposes, tests numerically and verifies experimentally a decentralized control algorithm with local feedback for semi-active mitigation of free vibrations in frame structures. The algorithm aims at transferring the vibration energy of low-order, lightly-damped structural modes into high-frequency modes of vibration, where it is quickly damped by natural mechanisms of material damping. Such an approach to mitigation of vibrations, known as the prestress–accumulation release (PAR) strategy, has been earlier applied only in global control schemes to the fundamental vibration mode of a cantilever beam. In contrast, the decentralization and local feedback allows the approach proposed here to be applied to more complex frame structures and vibration patterns, where the global control ceases to be intuitively obvious. The actuators (truss–frame nodes with controllable ability to transmit moments) are essentially unblockable hinges that become unblocked only for very short time periods in order to trigger local modal transfer of energy. The paper proposes a computationally simple model of the controllable nodes, specifies the control performance measure, yields basic characteristics of the optimum control, proposes the control algorithm and then tests it in numerical and experimental examples.

The modal transfer analysis by adding transport costs. Case study: The use of bus and private vehicle student Institut Teknologi Sumatera

Since the last two years the Institut Teknologi Sumatera to accept new students in large numbers. Increasing the number of students is directly proportional to the increase of private vehicles that enter the campus it is not in line with the policy of the Institut Teknologi Sumatera as Smart, Friendly, and Forest Campus. Institut Teknologi Sumatera as Smart, Friendly, and Forest Campus fact has made some deal with mass public transport, Damri, to provide bus transportation to the Institut Teknologi Sumatera, but the number of bus users is still inferior to the number of private vehicle users. This study was conducted to see the modal transfer of the entire academic community ITERA, from private cars to public transport such as buses, with the addition of transportation costs in the form of parking rates. This study shows the dominant displacement can occur if the respondent charged parking fees of IDR 4.000. Besides the displacement mode of transportation from using private vehicles be using the bus can occur if Trans Lampung fix three systems including bus departure and arrival schedules, facility service providers that support, and the addition of the Bus Trans Lampung.

The Impact of Resistance Training on Swimming Performance: A Systematic Review

The majority of propulsive forces in swimming are produced from the upper body, with strong correlations between upper body strength and sprint performance. There are significant gaps in the literature relating to the impact of resistance training on swimming performance, specifically the transfer to swimming performance. The aims of this systematic literature review are to (1) explore the transfer of resistance-training modalities to swimming performance, and (2) examine the effects of resistance training on technical aspects of swimming. Four online databases were searched with the following inclusion criteria: (1) journal articles with outcome measures related to swimming performance, and (2) competitive swimmers participating in a structured resistance-training programme. Exclusion criteria were (1) participants with a mean age <16years; (2) untrained, novice, masters and paraplegic swimmers; (3) triathletes and waterpolo players; (4) swimmers with injuries or illness; and (5) studies of starts and turns specifically. Data were extracted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, and the Physiotherapy Evidence Database (PEDro) scale was applied. For optimal transfer, specific, low-volume, high-velocity/force resistance-training programmes are optimal. Stroke length is best achieved through resistance training with low repetitions at a high velocity/force. Resisted swims are the most appropriate training modality for improving stroke rate. Future research is needed with respect to the effects of long-term resistance-training interventions on both technical parameters of swimming and overall swimming performance. The results of such work will be highly informative for the scientific community, coaches and athletes.

Efficient modeling of sound transmission through finite-sized thick and layered wall and floor systems

Built-up wall and floor systems such as roof panels, floors with floating screeds, etc., have found widespread application in building construction. Achieving sufficient sound insulation with these systems is challenging because of their relatively low weight and complex vibro-acoustic behavior. A fast and sufficiently accurate acoustic design tool is needed. The semi-analytical transfer matrix method is able to efficiently compute the response of a thick or multilayered structure in the frequency-wavenumber domain but has important limitations. First, the system is assumed to be of infinite extent. At lower frequencies however, neglecting the modal behavior of the wall can lead to large prediction errors. Second, integration over all possible incident plane waves is necessary to obtain the diffuse transmission loss, resulting in a high computation time. The transfer matrix approach is therefore extended in two ways. The modal behavior of rectangular walls and floors with simply supported boundary conditions is approximately accounted for. Using the diffuse reciprocity relationship, a hybrid modal transfer matrix-statistical energy analysis method is then developed such that integration of plane-wave transmission over all angles of incidence is no longer necessary, largely decreasing the computational effort. The model is validated against alternative numerical prediction models and experimental data.

Evidence for similar patterns of neural activity elicited by picture- and word-based representations of natural scenes

A long-standing core question in cognitive science is whether different modalities and representation types (pictures, words, sounds, etc.) access a common store of semantic information. Although different input types have been shown to activate a shared network of brain regions, this does not necessitate that there is a common representation, as the neurons in these regions could still differentially process the different modalities. However, multi-voxel pattern analysis can be used to assess whether, e.g., pictures and words evoke a similar pattern of activity, such that the patterns that separate categories in one modality transfer to the other. Prior work using this method has found support for a common code, but has two limitations: they have either only examined disparate categories (e.g. animals vs. tools) that are known to activate different brain regions, raising the possibility that the pattern separation and inferred similarity reflects only large scale differences between the categories or they have been limited to individual object representations. By using natural scene categories, we not only extend the current literature on cross-modal representations beyond objects, but also, because natural scene categories activate a common set of brain regions, we identify a more fine-grained (i.e. higher spatial resolution) common representation. Specifically, we studied picture- and word-based representations of natural scene stimuli from four different categories: beaches, cities, highways, and mountains. Participants passively viewed blocks of either phrases (e.g. "sandy beach") describing scenes or photographs from those same scene categories. To determine whether the phrases and pictures evoke a common code, we asked whether a classifier trained on one stimulus type (e.g. phrase stimuli) would transfer (i.e. cross-decode) to the other stimulus type (e.g. picture stimuli). The analysis revealed cross-decoding in the occipitotemporal, posterior parietal and frontal cortices. This similarity of neural activity patterns across the two input types, for categories that co-activate local brain regions, provides strong evidence of a common semantic code for pictures and words in the brain.

Enhanced vibration damping and modal energy transfer in a smart structure

The next challenge in the field of embedded structures is the development of an efficient, energetically autonomous vibration control technique. The proposed new approach consists in transferring energy between modes of the same structure: energy is harvested in a uncontrolled mode to increase attenuation in a targeted mode. This approach may even permit a ‘stand-alone’ design by generating constant voltages by using the structural motion with an energy harvesting technique. The proposed electrical architecture – called synchronized switch damping and harvesting (SSDH) – is composed of a harvesting circuit (synchronized switch harvesting on inductor [SSHI]), a DC-DC converter (buck-boost topology), and a modal circuit to control vibration (similar to combining the synchronized switch damping on voltage [SSDV] and modal synchronized switch damping on inductor [SSDI] methods). The theoretical analysis and numerical simulation of this architecture addressed the transfer of modal energy through a smart structure based on an academic case: a cantilever beam. The study aimed to increase damping in a selected mode by using the energy of other modes in configurations with realistic energy distributions. The circuit and energy transfer were analyzed to determine the minimum modal contributions necessary to achieve this aim and to ensure long-term stability (steady state). Compared with traditional SSDI, the bi-harmonic excitation used with this new technique increased attenuation damping by 13 dB. The proposed general approach can be extended to the majority of complex or composite industrial structures subjected to vibration.