No abstract

In practical applications of pattern recognition, there are often different features extracted from raw data which needs recognizing. Methods of combining multiple classifiers with different features are viewed as a general problem in various application areas of pattern recognition. In this paper, a systematic investigation has been made and possible solutions are classified into three frameworks, i.e. linear opinion pools, winner-take-all and evidential reasoning. For combining multiple classifiers with different features, a novel method is presented in the framework of linear opinion pools and a modified training algorithm for associative switch is also proposed in the framework of winner-take-all. In the framework of evidential reasoning, several typical methods are briefly reviewed for use. All aforementioned methods have already been applied to text-independent speaker identification. The simulations show that results yielded by the methods described in this paper are better than not only the individual classifiers' but also ones obtained by combining multiple classifiers with the same feature. It indicates that the use of combining multiple classifiers with different features is an effective way to attack the problem of text-independent speaker identification.

Forest aboveground biomass (AGB) was mapped throughout China using large footprint LiDAR waveform data from the Geoscience Laser Altimeter System (GLAS) onboard NASA's Ice, Cloud, and land Elevation Satellite (ICESat), Moderate Resolution Imaging Spectro-radiometer (MODIS) imagery and forest inventory data. The entire land of China was divided into seven zones according to the geographic characteristics of the forests. The forest AGB prediction models were separately developed for different forest types in each of the seven forest zones at GLAS footprint level from GLAS waveform parameters and biomass derived from height and diameter at breast height (DBH) field observation. Some waveform parameters used in the prediction models were able to reduce the effects of slope on biomass estimation. The models of GLAS-based biomass estimates were developed by using GLAS footprints with slopes less than 20° and slopes ≥ 20°, respectively. Then, all GLAS footprint biomass and MODIS data were used to establish Random Forest regression models for extrapolating footprint AGB to a nationwide scale. The total amount of estimated AGB in Chinese forests around 2006 was about 12,622 Mt vs. 12,617 Mt derived from the seventh national forest resource inventory data. Nearly half of all provinces showed a relative error (%) of less than 20%, and 80% of total provinces had relative errors less than 50%.

No abstract

We study how to learn a semantic parser of state-of-the-art accuracy with less supervised training data. We conduct our study on WikiSQL, the largest hand-annotated semantic parsing dataset to date. First, we demonstrate that question generation is an effective method that empowers us to learn a state-ofthe-art neural network based semantic parser with thirty percent of the supervised training data. Second, we show that applying question generation to the full supervised training data further improves the state-of-the-art model. In addition, we observe that there is a logarithmic relationship between the accuracy of a semantic parser and the amount of training data.

No abstract

Mapping the magnitude and spatial distribution of forest aboveground biomass (AGB, in Mg•ha −1) is crucial to improve our understanding of the terrestrial carbon cycle. Landsat/TM (Thematic Mapper) and ICESat/GLAS (Ice, Cloud, and land Elevation Satellite, Geoscience Laser Altimeter System) data were integrated to estimate the AGB in the Changbai Mountain area. Firstly, four forest types were delineated according to TM data classification. Secondly, different models for prediction of the AGB at the GLAS footprint level were developed from GLAS waveform metrics and the AGB was derived from field observations using multiple stepwise regression. Lastly, GLAS-derived AGB, in combination with vegetation indices, leaf area index (LAI), canopy closure, and digital elevation model (DEM), were used to drive a data fusion model based on the random forest approach for extrapolating the GLAS footprint AGB to a continuous AGB map. The classification result showed that the Changbai Mountain region was characterized as forest-rich in altitudinal vegetation zones. The contribution of remote sensing variables in modeling the AGB was evaluated. Vegetation index metrics account for large amount of contribution in AGB ranges <150 Mg•ha −1 , while canopy closure has the largest contribution in AGB ranges ≥150 Mg•ha −1. Our study revealed that spatial information from two sensors and DEM could be combined to estimate the AGB with an R 2 of 0.72 and an RMSE of 25.24 Mg•ha −1 in validation at stand level (size varied from~0.3 ha tõ 3 ha).

No abstract