Adaptive transfer learning framework for dense prediction of human activity recognition

Human activity recognition(HAR)for dense prediction is proven to be of good performance,but it relies on labeling every point in time series with the high cost.In addition,the performance of HAR model will show significant degradation when tested on the sensor data with different distribution from the training data,where the training data and the test data are usually collected from different sensor locations or sensor users.Therefore,the adaptive transfer learning framework for dense prediction of HAR is introduced to implement cross-domain transfer,where the proposed multi-level unsupervised domain adaptation(MLUDA)approach combines the global domain adaptation and the specific task adaptation to adapt the source and target domain in multiple levels.The multi-connected global domain adaptation architecture is proposed for the first time,which can adapt the output layer of the encoder and the decoder in dense prediction model.After this,the specific task adaptation is proposed to ensure alignment of each class centroid in source domain and target domain by introducing the cosine distance loss and the moving average method.Experiments on three public HAR datasets demonstrate that the proposed MLUDA improves the prediction accuracy of target data by 20%compared to the source domain pre-trained model and it is more effective than the other three deep transfer learning methods with an improvement of 10%to 18%in accuracy.

Preparation and Characterization of a Novel Benzimidazolium Bronsted Acid Ionic Liquid and Its Application in the Synthesis of Arylic Esters

A novel Brφnsted acid task specific ionic liquid 1-ethylbenzimidazolium tetrafluoroborate （[Hebim]BF4） with functional benzimidazolium cation was synthesized and characterized by ^1H NMR, IR, MS spectra and elemental analysis. This novel ionic liquid was successfully used as dual solvent-catalyst for the synthesis of arylic esters. Higher yields were obtained in the presence of [Hebim]BF4 in comparison with other imidazolium ionic liquids because of the good solubility of the aromatic alcohols and aromatic carboxylic acids in [Hebim]BF4. The product could be separated conveniently from the reaction system, and the ionic liquid could be easily reused after removal of water under vacuum. After 10 times reuse, the selectivity of the ester was still 100%.

Pebax/TSIL blend thin film composite membranes for CO_2 separation

In this study a thin film composite （TFC） membrane with a Pebax/Task-specific ionic liquid （TSIL） blend selective layer was prepared. Defect-flee Pebax/TSIL layers were coated successfully on a polysulfone ultrafiltration porous support with a poly- dimethylsiloxane （PDMS） gutter layer. Different parameters in the membrane preparation （e.g. concentration, coating time） were investigated and optimized. The morphology of the membranes was studied by scanning electron microscopy （SEM）, while the thermal properties and chemical structures of the membrane materials were investigated by thermo-gravimetric ana- lyzer （TGA）, differential scanning calorimetry （DSC） and Fourier transform infrared spectroscopy （FTIR）. The CO2 separation performance of the membrane was evaluated using a mixed gas permeation test. Experimental results show that the incorpora- tion of TSIL into the Pebax matrix can significantly increase both C02 permeance and CO2/N2 selectivity. With the presence of water vapor, the membrane exhibits the best CO2/N2 selectivity at a relative humidity of around 75%, where a CO2 permeance of around 500 GPU and a CO2/N2 selectivity of 46 were documented. A further increase in the relative humidity resulted in higher CO： permeance but decreased COIN2 selectivity. Experiments also show that CO2 permeance decreases with a CO2 partial pressure increase, which is considered a characteristic in facilitated transport membranes.

[bmim]OH： An efficient catalyst for the synthesis of mono and bis spirooxindole derivatives in ethanol at room temperature

A rapid and efficient, one pot synthesis of spirooxindole derivatives has been attempted by three- component reaction of isatin, malononitrile and carbonyl compound possessing a reactive c^-methylene group by using task specific ionic liquid, 1-butyl-3-methyl imidazolium hydroxide [bmim]OH as a catalyst. The important features of this methodology are straight forward route in short reaction time at room temperature and avoid any hazardous organic solvent, toxic catalyst, tedious purification step. Interestingly, this protocol is not only limited to mono-systems but also to the synthesis of newer bis- spirooxindole system. The separation of the product and reusability of the catalyst are easy with excellent yield. The [bmim]OH catalyst system could be reused up to five recycles without appreciable loss of activity.