incremental learning of triadic plsa for collaborative filtering

Beijing University of Technology (BJUT); Beijing Municipal Lab of Brain Informatics; Chinese Society of Radiology; National Natural Science Foundation of China (NSFC); State Administration of Foreign Experts Affairs

Visual-Context Boosting for Eye Detection

Eye detection plays an important role in many practical applications. This paper presents a novel two-step scheme for eye detection. The first step models an eye by a newly defined visual-context pattern (VCP), and the second step applies semisupervised boosting for precise detection. VCP describes both the space and appearance relations between an eye region (region of eye) and a reference region (region of reference). The context feature of a VCP is extracted by using the integral image. Aiming to reduce the human labeling efforts, we apply semisupervised boosting, which integrates the context feature and the Haar-like features for precise eye detection. Experimental results on several standard face data sets demonstrate that the proposed approach is effective, robust, and efficient. We finally show that this approach is ready for practical applications.

Context and conformation dictate function of a transcription antitermination switch

In bacteriophage, transcription elongation is regulated by the N protein, which binds a nascent mRNA hairpin ( termed boxB) and enables RNA polymerase to read through distal terminators. We have examined the structure, energetics and in vivo function of a number of N boxB complexes derived from in vitro protein selection. Trp18 fully stacks on the RNA loop in the wild-type structure, and can become partially or completely unstacked when the sequence context is changed three or four residues away, resulting in a recognition interface in which the best binding residues depend on the sequence context. Notably, in vivo antitermination activity correlates with the presence of a stacked aromatic residue at position 18, but not with N boxB binding affinity. Our work demonstrates that RNA polymerase responds to subtle conformational changes in cis-acting regulatory complexes and that approximation of components is not sufficient to generate a fully functional transcription switch.

A STUDY ON THE INDICATION OF UNEXPECTED COMPONENTS BY KALMAN FILTERING IN FLUOROSPECTROPHOTOMETRY

The potential of Kalman filtering for indication of unexpected components in a mixture was experimentally evaluated by taking the spectrofluorimetric analysis of the tricomponent system;oi phenylalanine, tryhtophen and tyrosine as an example. According to

COMPARATIVE-STUDY OF KALMAN FILTERING, SYNCHRONOUS EXCITATION AND NUMERICAL DERIVATIVE TECHNIQUES IN FLUOROMETRY

The performance of Kalman filtering, synchronous excitation and numerical derivative techniques for the resolution of overlapping emission spectra in spectrofluorimetry was studied. The extent of spectrum overlap was quantitatively described by the separation degree D(s), defined as the ratio of the peak separation to the full width at half-maximum of the emission spectrum of the interferent. For the system of Rhodamine B and Rhodamine 6G with a large D(s) of about 0.4, both Kalman filtering and synchronous techniques are able to resolve the overlapping spectra well and to give satisfactory results while the derivative spectra are still overlapped with each other. Moreover, the sensitivities are greatly decreased in derivative techniques. For more closely spaced spectra emitted by the complexes of Al and Zn with 7-iodo-8-hydroxyquinoline-5-sulphonic acid (Ferron)-hexadecyltrimethylammonium bromide, the synchronous excitation technique cannot completely separate the overlapping peaks, although it increases the separation degree from 0.25 in the conventional spectra to 0.37 in the synchronous spectra. On the other hand, Kalman filtering is capable of resolving this system. When the Al/Zn intensity ratio at the central wavelength of Al was > 1, however, the accuracy and precision of the estimates for Zn concentration produced by the Kalman filter became worse. In this event, the combination of synchronous excitation and Kalman filtering can much improve the analytical results.

RESOLUTION OF CLOSELY SPACED LINES WITH KALMAN FILTERING IN INDUCTIVELY COUPLED PLASMA ATOMIC EMISSION-SPECTROMETRY

Effects of some factors on the performance of our Kalman filter in discrimination of closely spaced overlapping signals were investigated. The resolution power of the filter for overlapping lines can be strengthened by reduction of the step size in scans. The minimum peak separation of two lines which the Kalman filter can effectively handle generally equals two to three times the step size in scans. Significant difference between the profiles of the analysis and interfering lines and multiple lines from matrix in the spectral window of the analysis line are very helpful for the Kalman filter to discern closely spaced analysis and interfering signals correctly, which allow the filter well to resolve the line pair with very small peak distance or even the entirely coincident lines.

KALMAN FILTERING FOR RESOLUTION OF OVERLAPPING LINES IN INDUCTIVELY COUPLED PLASMA ATOMIC EMISSION-SPECTROMETRY

A Kalman filter was developed for resolving overlapping lines in inductively coupled plasma atomic emission spectrometry (ICP-AES) and evaluated experimentally with the determination of La in the presence of Ho, and Cu in the presence of Pr. The whiteness of the innovation sequence for an optimal filter was explored to be the criterion for the correction of the wavelength positioning errors which may occur in spectral scans. Under the conditions of the medium-resolution spectrometer and 1.5 pm step size in scans, the filter effectively resolved the Cu/Pr line pair having a small peak separation of 4.8 pm. For the La/Ho line pair with a peak distance of 9.8 pm, an unbiased estimate for La concentration was still obtained even when the signal-to-background ratio was down to 0.048. Favourable detection limits for real samples were achieved. Unstructured backgrounds were modeled theoretically and all spectral scans therefore did not require the correction for solvent.

EVALUATION OF THE RELIABILITY OF KALMAN FILTERING RESULTS IN INDUCTIVELY COUPLED PLASMA ATOMIC EMISSION-SPECTROMETRY

This paper deals with the evaluation of the reliability of the analytical results obtained by Kalman filtering. Two criteria for evaluation were compared: one is based on the autocorrelation analysis of the innovation sequence, the so-called NAC criterion; the other is the innovations number, which actually is the autocorrelation coefficient of the innovation sequence at the initial wavelength. Both criteria allow compensation for the wavelength positioning errors in spectral scans, but there exists a difference in the way they work. The NAC criterion can provide information about the reliability of an individual result, which is very useful for the indication of unmodelled emissions, while the innovations number should be incorporated with the normalization of the innovations or seek the help of the sequence itself for the same purpose. The major limitation of the NAC criterion is that it does not allow the theoretical modelling of continuous backgrounds, which, however, is convenient in practical analysis and can be taken with the innovations number criterion.

EVALUATION OF THE EFFECT OF WAVELENGTH POSITIONING ERRORS ON KALMAN FILTERING RESULTS IN INDUCTIVELY COUPLED PLASMA ATOMIC EMISSION-SPECTROMETRY

This work evaluates the effect of wavelength positioning errors in spectral scans on analytical results when the Kalman filtering technique is used for the correction of line interferences in inductively coupled plasma atomic emission spectrometry (ICP-AES). The results show that a positioning accuracy of 0.1 pm is required in order to obtain accurate and precise estimates for analyte concentrations. The positioning error in sample scans is more crucial than that in model scans. The relative bias in measured analyte concentration originating from a positioning error in a sample scan increases linearly with an increase in the magnitude of the error and the peak distance of the overlapping lines, but is inversely proportional to the signal-to-background ratio. By the use of an optimization procedure for the positions of scans with the innovations number as the criterion, the wavelength positioning error can be reduced and, correspondingly, the accuracy and precision of analytical results improved.

Density-dependent effects on seston dynamics and rates of filtering and biodeposition of the suspension-cultured scallop Chlamys farreri in a eutrophic bay (northern China): An experimental study in semi-in situ flow-through systems

Effects of stocking density on seston dynamics and filtering and biodeposition by the suspension-cultured Zhikong scallop Chlamys farreri Jones et Preston in a eutrophic bay (Sishili Bay, northern China), were determined in a 3-month semi-field experiment with continuous flow-through seawater from the bay. Results showed that the presence of the scallops could strongly decrease seston and chlorophyll a concentrations in the water column. Moreover, in a limited water column, increasing scallop density could cause seston depletion due to scallop's filtering and biodeposition process, and impair scallop growth. Both filtration rate and biodeposition rate of C. farreri showed significant negative correlation with their density and positive relationship with seston concentration. Calculation predicts that the daily removal of suspended matter from water column by the scallops in Sishili Bay ecosystem can be as high as 45% of the total suspended matter; and the daily production of biodeposits by the scallops in early summer in farming zone may amount to 7.78 g m(-2), with daily C, N and P biodeposition rates of 3.06 x 10(-1), 3.86 x 10(-2) and 9.80 x 10(-3) g m(-2), respectively. The filtering and biodeposition by suspension-cultured scallops could substantially enhance the deposition of total suspended particulate material, suppress accumulation of particulate organic matter in water column, and increase the flux of C, N and P to benthos, strongly enhancing pelagic-benthic coupling. It was suggested that the filtering-biodeposition process by intensively suspension-cultured bivalve filter-feeders could exert strong top-down control on phytoplankton biomass and other suspended particulate material in coastal ecosystems. This study also indicated that commercially suspension-cultured bivalves may simultaneously and potentially aid in mitigating eutrophication pressures on coastal ecosystems subject to anthropogenic N and P loadings, serving as a eutrophic-environment bioremediator. The ecological services (e.g. filtering capacity, top-down control, and benthic-pelagic coupling) functioned by extractive bivalve aquaculture should be emphasized in coastal ecosystems. (c) 2005 Elsevier B.V. All rights reserved.

Influence of filtering and biodeposition by the cultured scallop Chlamys farreri on benthic-pelagic coupling in a eutrophic bay in China

In coastal ecosystems, suspension-cultured bivalve filter feeders may exert a strong impact on phytoplankton and other suspended particulate matter and induce strong pelagic-benthic coupling via intense filtering and biodeposition. We designed an in situ method to determine spatial variations in the filtering-biodeposition process by intensively suspension-cultured scallops Chlamys farreri in summer in a eutrophic bay (Sishili Bay, China), using cylindrical biodeposition traps directly suspended from longlines under ambient environmental conditions. Results showed that bivalve filtering-biodeposition could substantially enhance the deposition of total suspended material and the flux of C, N and P to the benthos, indicating that the suspended filter feeders could strongly enhance pelagic-benthic coupling and exert basin-scale impacts in the Sishili Bay ecosystem. The biodeposition rates of 1-yr-old scallops varied markedly among culture sites (33.8 to 133.0 mg dry material ind.(-1) d(-1)), and were positively correlated with seston concentrations. Mean C, N and P biodeposition rates were 4.00, 0.51, 0.11 mg ind.-1 d-1, respectively. The biodeposition rates of 2-yr-old scallops were almost double these values. Sedimentation rates at scallop culture sites averaged 2.46 times that at the reference site. Theoretically, the total water column of the bay could be filtered by the cultured scallops in 12 d, with daily seston removal amounting to 64%. This study indicated that filtering-biodeposition by suspension-cultured scallops could exert long-lasting top-down control on phytoplankton biomass and other suspended material in the Sishili Bay ecosystem. In coastal waters subject to anthropogenic N and P inputs, suspended bivalve aquaculture could be advantageous, not only economically, but also ecologically, by functioning as a biofilter and potentially mitigating eutrophication pressures. Compared with distribution-restricted wild bivalves, suspension-cultured bivalves in deeper coastal bays may be more efficient in processing seston on a basin scale.

Generalized Low-Rank Approximations

We study the frequent problem of approximating a target matrix with a matrix of lower rank. We provide a simple and efficient (EM) algorithm for solving {\\em weighted} low rank approximation problems, which, unlike simple matrix factorization problems, do not admit a closed form solution in general. We analyze, in addition, the nature of locally optimal solutions that arise in this context, demonstrate the utility of accommodating the weights in reconstructing the underlying low rank representation, and extend the formulation to non-Gaussian noise models such as classification (collaborative filtering).

Context-Based Vision System for Place and Object Recognition

While navigating in an environment, a vision system has to be able to recognize where it is and what the main objects in the scene are. In this paper we present a context-based vision system for place and object recognition. The goal is to identify familiar locations (e.g., office 610, conference room 941, Main Street), to categorize new environments (office, corridor, street) and to use that information to provide contextual priors for object recognition (e.g., table, chair, car, computer). We present a low-dimensional global image representation that provides relevant information for place recognition and categorization, and how such contextual information introduces strong priors that simplify object recognition. We have trained the system to recognize over 60 locations (indoors and outdoors) and to suggest the presence and locations of more than 20 different object types. The algorithm has been integrated into a mobile system that provides real-time feedback to the user.