Jiun-Wei Liou

Autoencoder for words

This paper presents a training method that encodes each word into a different vector in semantic space and its relation to low entropy coding. Elman network is employed in the method to process word sequences from literary works. The trained codes possess reduced entropy and are used in ranking, indexing, and categorizing literary works. A modification of the method to train the multi-vector for each polysemous word is also presented where each vector represents a different meaning of its word. These multiple vectors can accommodate several different meanings of their word. This method is applied to the stylish analyses of two Chinese novels, Dream of the Red Chamber and Romance of the Three Kingdoms.

Construct adaptive template array for magnetic resonance images

Dealing with large number of brain images in group analysis involves two kinds of analysis. One is to extract the information and relations in the population of brains, and the other is to combine the information of individual brain in the group. Linear or nonlinear dimension reduction algorithms are the main tools to perform the first analysis, which is to show the information of the population. The hidden relations in the distribution are therefore able to be visualized in low-dimensional and visible space. Image registration is the critical part of the second analysis, which is to integrate the information or statistics of individual brains. The statistics are registered to a template which is commonly the mean brain image of the population so that the statistics from different subjects can be compared in the same stereotaxic space. The process of registering images to the template is called normalization. The quality of registration decides the normalization and the interpretability of results. This work constructed ordered representations, from a set of brain images, as the multiple templates. The ordered representations are derived from self-organizing map. A novel method, transformation diversion, based on the ordered representations is proposed to improve the registration, which is a non-linear deformation, in a general manner. The discriminative low-dimensional representation of the population of Alzheimer disease and normal subjects are also shown. The set of ordered representations not only shows the population information but also improve the normalization process.

Intensity Gradient Self-organizing Map for Cerebral Cortex Reconstruction

This paper presents an application of a self-organizing map (SOM) model based on the image intensity gradient for the reconstruction of cerebral cortex from MR images. The cerebral cortex reconstruction is important for many brain science or medicine related researches. However, it is difficult to extract deep cortical folds. In our method, we apply the SOM model based on the image intensity gradient to deform the easily extracted white matter surface and extract the cortical surface. The intensity gradient vectors are calculated according to the intensities of image data. Thus the proper cortical surface can be extracted from the image information itself but not artificial features. The simulations on T1-weighted MR images show that the proposed method is robust to reconstruct the cerebral cortex.

Reliability Maps in Event Related Functional MRI Experiments

In functional magnetic resonance imaging (fMRI) studies, the blood oxygen level-dependent (BOLD) signal change, in contrast to noise, is typically small (< 5%; e.g., Chen & Small, 2007). Although the quality of acquired image data may be improved by pre-processing images with lowor high-pass filters, classification of voxels into the active/inactive status could vary from one study to the next even when the same experimental paradigm is implemented (Maitra, 2009). Reliability assessment would contribute significantly to the knowledge on noise structures in image data, as a function of stimulus sequences, ethnic groups, imaging techniques and scanner differences (Biswal et al., 1996; Genovese et al., 1997; Maitra et al., 2002).

Structure sensitive complexity for symbol-free sequences

The study proposes our extended method to assess structure complexity for symbol-free sequences, such as literal texts, DNA sequences, rhythm, and musical input. This method is based on L-system and topological entropy for context-free grammar. Inputs are represented as binary trees. Different input features are represented separately within tree structure and actual node contents. Our method infers tree generating grammar and estimates its complexity. This study reviews our previous results on texts and DNA sequences and provides new information regarding them. Also, we show new results measuring complexity of Chinese classical texts and music samples with rhythm and melody components. Our method demonstrates enough sensitivity to extract quasi-regular structured fragments of Chinese texts and to detect irregular styled samples of music inputs. To our knowledge, there is no other method that can detect such quasi-regular patterns.

About eigenvalues from embedding data complex in low dimension

LLE(Local linear embedding) and Isomap are widely used approaches for dimension reduction on data complex. The embedding results from the two methods are eigenvectors from solving specific matrices. The corresponding eigenvalues for the selected eigenvectors have important meaning for the embedding results. In this paper, the k-nn method and e-distance approach are used for neighborhood function with parameters. Then, different datasets and parameters will be applied to obtain the embedding results and eigenvalues. The main change of eigenvalues and the corresponding embedding results will be shown in this paper.

Neighborhood selection and eigenvalues for embedding data complex in low dimension

LLE(Local linear embedding) and Isomap are widely used approaches for dimension reduction. For LLE, the neighborhood selection approach is an important research issue. For different types of datasets, we need different neighborhood selection approaches to have better chance for finding reasonable representation within the required number of dimensions. In this paper, the e-distance approach and a modified version of k-nn method are introduced. For LLE and Isomap, the eigenvectors obtained from these methods are much more discussed, but there are more information hidden in the corresponding eigenvalues which can be used for finding embeddings contains more data information.

Self-Organizing reinforcement learning model

A motor control model based on reinforcement learning (RL) is proposed here. The model is inspired by organizational principles of the cerebral cortex, specifically on cortical maps and functional hierarchy in sensory and motor areas of the brain. Self-Organizing Maps (SOM) have proven to be useful in modeling cortical topological maps. The SOM maps the input space in response to the real-valued state information, and a second SOM is used to represent the action space. We use the Q-learning algorithm with a neighborhood update function, and an SOM for Q-function to avoid representing very large number of states or continuous action space in a large tabular form. The final model can map a continuous input space to a continuous action space.