Hirotaka Niitsuma

Biologically motivated vergence control system using human-like selective attention model

We propose a new human-like vergence control method for an active stereo vision system. The proposed system uses a selective attention model to localize an interesting area in each camera. The selected object area in the master camera is compared with that in the slave camera to identify whether the two cameras find a same landmark. If the left and right cameras successfully find a same landmark, the implemented active vision system with two cameras focuses on the landmark. Using the motor encoder information, we can detect depth information automatically. Computer simulation and experimental results show that the proposed vergence control method is very effective in implementing the human-like active stereo vision system.

Covariance and PCA for categorical variables

Covariances from categorical variables are defined using a regular simplex expression for categories. The method follows the variance definition by Gini, and it gives the covariance as a solution of simultaneous equations using the Newton method. The calculated results give reasonable values for test data. A method of principal component analysis (RS-PCA) is also proposed using regular simplex expressions, which allows easy interpretation of the principal components.

λ-Opt Neural Approaches to Quadratic Assignment Problems

In this article, we propose new analog neural approaches to combinatorial optimization problems, in particular, quadratic assignment problems (QAPs). Our proposed methods are based on an analog version of the -opt heuristics, which simultaneously changes assignments for elements in a permutation. Since we can take a relatively large value, our new methods can achieve a middle-range search over possible solutions, and this helps the system neglect shallow local minima and escape from local minima. In experiments, we have applied our methods to relatively large-scale (N = 80150) QAPs. Results have shown that our new methods are comparable to the present champion algorithms; for two benchmark problems, they are obtain better solutions than the previous champion algorithms.

Spiral mining using attributes from 3d molecular structures

Active responses from analysts play an essential role in obtaining insights into structure activity relationships (SAR) from drug data. Experts often think of hypotheses, and they want to reflect these ideas in the attribute generation and selection process. We analyzed the SAR of dopamine agonists and antagonists using the cascade model. The presence or absence of linear fragments in molecules constitutes the core attribute in the mining. In this paper, we generated attributes indicating the presence of hydrogen bonds from 3D coordinates of molecules. Various improvements in the fragment expressions are also introduced following the suggestions of chemists. Attribute selection from the generated fragments is another key step in mining. Close interactions between chemists and system developers have enabled spiral mining, in which the analysis results are incorporated into the development of new functions in the mining system. All these factors are necessary for success in SAR mining.

A NON-PARAMETRIC TRAINABLE OBJECT-DETECTION MODEL USING A CONCEPT OF RETINOTOPIC SAMPLING

A retina has a space-variant sampling mechanism and an orientation-sensitive mechanism. The space-variant sampling mechanism of the retina is called retinotopic sampling (RS).Withthese mechanisms of the retina, the object-detection isformulated as finding appropriatecoordinate transformation from a coordinate system on an input image, to a coordinate system on the retina. However, when the object size is inferred by thismechanism, the result tends to gravitate towards zero. To cancel this gravity, the space-variant sampling mechanism is modified to uniform sampling mechanism, but a concept of RS is equivalently introduced by using space-variant weights. This object-detection mechanism is modeled as a non-parametric method. By using the model based on RS, we formulate a kernel function as an analytical function of information of an object, a position and a size of the object in an image. Then the object-detection is realized as a gradient decent method for a discriminant function trained by Support Vector Machine (SVM) using this kernel function. This detection mechanism realizes faster detection than exploring a visual scene in raster-like fashion. The discriminant function outperforms results of SVMs using a kernel function using intensities of all pixels (based on independently published results), in face detection experiments over the 24,045 test images in the MIT-CBCL database.