Domain Adaptation for Structured Regression

Discriminative regression models have proved effective for many vision applications (here we focus on 3D full-body and head pose estimation from image and depth data). However, dataset bias is common and is able to significantly degrade the performance of a trained model on target test sets. As we show, covariate shift, a form of unsupervised domain adaptation (USDA), can be used to address certain biases in this setting, but is unable to deal with more severe structural biases in the data. We propose an effective and efficient semi-supervised domain adaptation (SSDA) approach for addressing such more severe biases in the data. Proposed SSDA is a generalization of USDA, that is able to effectively leverage labeled data in the target domain when available. Our method amounts to projecting input features into a higher dimensional space (by construction well suited for domain adaptation) and estimating weights for the training samples based on the ratio of test and train marginals in that space. The resulting augmented weighted samples can then be used to learn a model of choice, alleviating the problems of bias in the data; as an example, we introduce SSDA twin Gaussian process regression (SSDA-TGP) model. With this model we also address the issue of data sharing, where we are able to leverage samples from certain activities (e.g., walking, jogging) to improve predictive performance on very different activities (e.g., boxing). In addition, we analyze the relationship between domain similarity and effectiveness of proposed USDA versus SSDA methods. Moreover, we propose a computationally efficient alternative to TGP (Bo and Sminchisescu 2010), and it’s variants, called the direct TGP. We show that our model outperforms a number of baselines, on two public datasets: HumanEva and ETH Face Pose Range Image Dataset. We can also achieve 8–15 times speedup in computation time, over the traditional formulation of TGP, using the proposed direct formulation, with little to no loss in performance.     

Development of large-scale stacked-type electrostatic actuators for use as artificial muscles

Electrostatic actuators have the advantages of light weight, flexibility, and high energy efficiency, which make them suitable for use as artificial muscles. However, a traditional electrostatic actuator cannot generate long strokes and a high force density at the same time because such actuator would excessively widen the gap between the electrodes because of its structure. This paper presents a newly developed large-scale stacked-type electrostatic actuator (LSEA) intended for use as an artificial muscle for robots. LSEA is a multi-stacked electrostatic actuator that can be linearly contracted by the application of a voltage. It has a unique structure that prevents overextension of the gap between the electrodes. It can therefore generate a large force. The spring characteristics and the relationship between the contractive force and the stroke were experimentally determined. The findings showed that LSEA prevents the overextension of the gap between the electrodes and has a high contraction ratio that is equivalent to that of a mammalian skeletal muscle.     

Fate of the chlorine and fluorine in a sub-bituminous coal during pyrolysis and gasification

The fate of the chlorine and fluorine present in a sub-bituminous coal from Indonesia during pyrolysis and gasification has been studied with fixed and entrained bed reactors. The rate profile for HCl evolved in the temperature programmed pyrolysis exhibits the main and shoulder peaks at 480 and 600 °C, respectively. Model experiments and subsequent Cl 2p XPS measurements show that HCl reacts with metal impurities and carbon active sites at 500 °C to be retained as inorganic and organic chlorine forms, from which HCl evolves again at elevated temperatures. It is suggested that the HCl observed in the coal pyrolysis may originate from the above-mentioned chlorine functionalities formed by secondary reactions involving the nascent char. In the CO₂ gasification of the 900 °C char at 1000 °C and 2.5 MPa, any measurable amounts of HCl and HF could not be detected even at a high conversion of 75 wt% (daf), suggesting the accumulation of these halogens in the residual char. When the coal is injected into an O₂-blown, entrained bed gasifier at 1200-1400 °C under 2.6 MPa, the partial oxidation proceeds to a larger extent at a higher O₂/coal ratio, whereas the chlorine and fluorine are enriched in the remaining char, and the extent of the enrichment at the latter stage of gasification is larger with the fluorine. The XPS measurements of the chars reveal the presence of the broad F 1 s peak, which can cover a wide range of binding energies attributable to inorganic and organic fluorine. The halogen enrichment during gasification is discussed in terms of secondary reactions of HCl and HF with char.     

Two-stage pyrolysis of heavy oils. 3. Pyrolysis of tar-sand bitumens. Relation between ethylene yield and structural characteristics of heavy oils

Thermal cracking of tar-sand bitumens has been carried out using a two-stage pyrolysis reactor with temperature zones of 440°C and 750–800°C, respectively. Feedstocks were pyrolysed in the first stage into cracked oils, which were carried to the second stage for subsequent pyrolysis. Only 12–14 wt% of ethylene was obtained from tar-sand bitumens at the residence time of 1.2 s in the second stage, although 27 and 16 wt% were obtained from Taching and Iranian heavy vacuum residues, respectively. The tar-sand bitumens contain shorter paraffinic straight-chains and have more branched molecules than the vacuum residues of petroleum. A straight-chain paraffin index is proposed, with which a good correlation was obtained between ethylene yield and the fraction of straight-chain paraffin carbons in the heavy oil.     

Nitrous oxide emissions from fertilized upland fields in Thailand

Nitrous oxide (N₂O) emission from fertilized maize fields was measured using a closed chamber at four experimental sites in Thailand. The average measured N₂O flux from unfertilized plots through crop season was 4.16 ± 1.52, 5.05 ± 1.65, 5.25 ± 1.68 and 6.74 ± 2.95 g N₂O-N m^-2 h^-1, at Nakhon Sawan, Phra Phutthabat, Khon Kaen and Chiang Mai, respectively. Increased N₂O emissions by the application of nitrogen fertilizer were 0.22–0.44, 0.19–0.38%, 0.12–0.24 and 0.08–0.15% of the applied N, respectively. Compared to other data, N₂O emission rate to applied nitrogen was not significantly different between the data of Thailand and the Temperate Zone.     

The Direct Synthesis of β-hydroxy-acids by Lithium Naphthalene and Acetic Acid

Lithium naphthalene reacts with acetic acid to give α-anion of lithium acetate. The reaction of this anion with various ketones gives the corresponding β-hydroxy-acids in good yield.     

Performance of solvothermally prepared Ga2O3-Al2O3 catalysts for SCR of NO with methane

The solvothermal reaction of mixtures of aluminum isopropoxide (AIP) and gallium acetylacetonate (Ga(acac)₃) directly yielded the mixed oxides of γ-Ga₂O₃-Al₂O₃. In the solvothermal synthesis, the crystal structure of mixed oxides was controlled by the initial formation of γ-Ga₂O₃ nuclei. The mixed oxides prepared in diethylenetriamine have extremely high activities for selective catalytic reduction (SCR) of NO with methane as a reducing agent. With increasing crystallite size of the spinel structure, the catalytic activity increased. The ratio of the amount of methane consumed by combustion to total methane conversion was proportional to the density of acid sites on the surface of the mixed oxides. The mixed oxide catalysts prepared in diethylenetriamine had lower densities of acid sites and showed a higher methane-efficiency for CH₄-SCR than those prepared in other solvents. These catalysts maintained their high activity even when the reaction was carried out under the severe conditions (i.e., high space velocity and low NO concentration).     

Observation of single-walled carbon nanotubes by photoemission microscopy

Single-walled carbon nanotube networks grown on SiO₂ pillars were studied by means of scanning photoemission microscopy. The individual nanotubes or nanotube bundles growing from the pillar tops were observed in C 1s images. Band bending near catalytic Fe/nanotube contacts in an end-bonded configuration was studied by measuring C 1s spectra along the tube axes. Within our experimental resolution, no band bending was observed. This implies that the depletion width is less than the spatial resolution of the scanning photoemission microscope (90 nm) or that the amount of the band bending is less than 0.1 eV.