No abstract

This paper proposes a new brain-computer interface (BCI) design using fuzzy ARTMAP (FA) neural network, as well as an application of the design. The objective of this BCI-FA design is to classify the best three of the five available mental tasks for each subject using power spectral density (PSD) values of electroencephalogram (EEG) signals. These PSD values are extracted using the Wiener-Khinchine and autoregressive methods. Ten experiments employing different triplets of mental tasks are studied for each subject. The findings show that the average BCI-FA outputs for four subjects gave less than 6% of error using the best triplets of mental tasks identified from the classification performances of FA. This implies that the BCI-FA can be successfully used with a tri-state switching device. As an application, a proposed tri-state Morse code scheme could be utilized to translate the outputs of this BCI-FA design into English letters. In this scheme, the three BCI-FA outputs correspond to a dot and a dash, which are the two basic Morse code alphabets and a space to denote the end (or beginning) of a dot or a dash. The construction of English letters using this tri-state Morse code scheme is determined only by the sequence of mental tasks and is independent of the time duration of each mental task. This is especially useful for constructing letters that are represented as multiple dots or dashes. This combination of BCI-FA design and the tri-state Morse code scheme could be developed as a communication system for paralyzed patients.

When viewing a painting, artists perceive more information from the painting on the basis of their experience and knowledge than art novices do. This difference can be reflected in eye scan paths during viewing of paintings. Distributions of scan paths of artists are different from those of novices even when the paintings contain no figurative object (i.e. abstract paintings). There are two possible explanations for this difference of scan paths. One is that artists have high sensitivity to high-level features such as textures and composition of colors and therefore their fixations are more driven by such features compared with novices. The other is that fixations of artists are more attracted by salient features than those of novices and the fixations are driven by low-level features. To test these, we measured eye fixations of artists and novices during the free viewing of various abstract paintings and compared the distribution of their fixations for each painting with a topological attentional map that quantifies the conspicuity of low-level features in the painting (i.e. saliency map). We found that the fixation distribution of artists was more distinguishable from the saliency map than that of novices. This difference indicates that fixations of artists are less driven by low-level features than those of novices. Our result suggests that artists may extract visual information from paintings based on high-level features. This ability of artists may be associated with artists’ deep aesthetic appreciation of paintings.

The human brain can effectively learn a new task from a small number of samples, which indicates that the brain can transfer its prior knowledge to solve tasks in different domains. This function is analogous to transfer learning (TL) in the field of machine learning. TL uses a well-trained feature space in a specific task domain to improve performance in new tasks with insufficient training data. TL with rich feature representations, such as features of convolutional neural networks (CNNs), shows high generalization ability across different task domains. However, such TL is still insufficient in making machine learning attain generalization ability comparable to that of the human brain. To examine if the internal representation of the brain could be used to achieve more efficient TL, we introduce a method for TL mediated by human brains. Our method transforms feature representations of audiovisual inputs in CNNs into those in activation patterns of individual brains via their association learned ahead using measured brain responses. Then, to estimate labels reflecting human cognition and behavior induced by the audiovisual inputs, the transformed representations are used for TL. We demonstrate that our brain-mediated TL (BTL) shows higher performance in the label estimation than the standard TL. In addition, we illustrate that the estimations mediated by different brains vary from brain to brain, and the variability reflects the individual variability in perception. Thus, our BTL provides a framework to improve the generalization ability of machine-learning feature representations and enable machine learning to estimate human-like cognition and behavior, including individual variability.

No abstract

Recent evidence has demonstrated that spatiotemporal patterns of spontaneous activity reflect the patterns of activity evoked by sensory stimuli. However, few studies have examined whether response profiles of task-evoked activity, which is not related to external sensory stimuli but rather to internal processes, are also reflected in those of spontaneous activity. To address this, we recorded activity of neurons in the lateral intraparietal area (LIP) when monkeys performed reaction-time and delayed-response visual-search tasks. We particularly focused on the target location-dependent modulation of delay-period activity (delay-period modulation) in the delayed-response task, and the discharge-rate persistency in fixation-period activity (baseline-activity maintenance) in the reaction-time task. Baseline-activity maintenance was assessed by the correlation between the spike counts of 2 separate bins. We found that baseline-activity maintenance, calculated from bins separated by a long interval (200-500 ms), was correlated with delay-period modulation, whereas that calculated from bins separated by a short interval (~100 ms) was correlated with trial-to-trial fluctuations in baseline activity, suggesting a link between the capability to hold task-related information in delay-period activity and the degree of baseline-activity maintenance in a timescale-dependent manner.

Many recommender systems employed in commercial web sites use collaborative filtering. The main goal of traditional collaborative filtering techniques is improvement of the accuracy of recommendation. Nevertheless, such techniques present the problem that they include many items that the user already knows. These recommendations appear to be good when we consider accuracy alone. On the other hand, when we consider users' satisfaction, they are not necessarily good because of the lack of discovery. In our work, we infer items that a user does not know by calculating the similarity of users or items based on information about what items users already know. We seek to recommend items that the user would probably like and does not know by combining the above method and the most popular method of collaborative filtering.

Information filtering systems, which recommend appropriate information to users from enormous amount of information, are becoming popular. One method of information filtering is content-based filtering that compares a user profile with a content model. Many systems using content-based filtering deal with text data, and few systems deal with music data. We propose a content-based filtering system for music data by using a decision tree. Compared with other filtering methods, a decision tree can eliminate noise features, which are not related to the user's preference, and can allow the user to edit the learned user profile. We conduct an experiment by using real music data and users to validate the effectiveness of our system compared with other filtering methods.