Yukihiro Tsuboshita

Gender estimation for SNS user profiling using automatic image annotation

User profiling for Social Network Services (SNS) has gained great attention because of its potential values in identifying target population, which is very informative for marketing. Many studies have been conducted to estimate SNS user profiles using text analysis. However, in spite of the huge quantities of image resources on SNS, no previous work has specifically explored user profiles by automatic image annotation techniques. This paper addresses the problem of inferring a SNS users gender by automatic image annotation. The proposed method involves learning a model to annotate SNS images and integrating annotation scores of images to infer a users gender. Evaluation based on Twitter data demonstrates promising results.

Context-dependent retrieval of information by neural-network dynamics with continuous attractors

Memory retrieval in neural networks has traditionally been described by dynamic systems with discrete attractors. However, recent neurophysiological findings of graded persistent activity suggest that memory retrieval in the brain is more likely to be described by dynamic systems with continuous attractors. To explore what sort of information processing is achieved by continuous-attractor dynamics, keyword extraction from documents by a network of bistable neurons, which gives robust continuous attractors, is examined. Given an associative network of terms, a continuous attractor led by propagation of neuronal activation in this network appears to represent keywords that express underlying meaning of a document encoded in the initial state of the network-activation pattern. A dominant hypothesis in cognitive psychology is that long-term memory is archived in the network structure, which resembles associative networks of terms. Our results suggest that keyword extraction by the neural-network dynamics with continuous attractors might symbolically represent context-dependent retrieval of short-term memory from long-term memory in the brain.

Content-Aware Multi-task Neural Networks for User Gender Inference Based on Social Media Images

To estimate demographic attributes such as gender and age of social media users from images posted by the users is a challenging problem because the demographic attributes are directly not shown in images. For such problem, prior approaches can be roughly separated into two types: one approach uses concept detection to detect pre-defined visual concepts which are then used as meta-data to estimate demographic attributes and the other approach directly uses content features such as Fisher Vector [19] which are extracted from images. In this paper we consider the way of combining these two approaches. We propose Multi-task Bilinear Model for integrating the detected concepts with the content features. In our proposed method, both the concept detector and the feature extractor can be jointly learned with end-to-end fashion. We evaluated the proposed method for the task of estimating user gender from Twitter images and found that it outperformed other baseline methods.

Information retrieval based on a neural-network system with multi-stable neurons

Neurophysiological findings of graded persistent activity suggest that memory retrieval in the brain is described by dynamical systems with continuous attractors. It has recently been shown that robust graded persistent activity is generated in single cells. Multiple levels of stable activity at a single cell can be replicated by a model neuron with multiple hysteretic compartments. Here we propose a framework to simply calculate the dynamical behavior of a network of multi-stable neurons. We applied this framework to spreading activation for document retrieval. Our method shows higher performance of retrieval than other spreading activation methods. The present study thus presents novel and useful information-processing algorithm inferred from neuroscience.

A Weighted Combination of Text and Image Classifiers for User Gender Inference

Demographic attribute inference of social networking service (SNS) users is a valuable application for marketing and for targeting advertisements. Several studies have examined Twitter-user gender inference in natural language processing, image recognition, and other research domains. Reportedly, a combined approach using text data and image data outperforms an individual data approach. This paper presents a proposal of a novel hybrid approach. A salient benefit of our system is that features provided from a text classifier and from an image classifier are combined appropriately to infer male or female gender using logistic regression. The experimentally obtained results demonstrate that our approach markedly improves an existing combination-based method.

Statistical–Mechanical Analysis of Attractor Dynamics in a Hysteretic Neuron Network

The attractor dynamics of a neural network of neurons with a hysteretic response property are investigated by a statistical–mechanical approach and numerical simulations using the Hopfield associative-memory model. Exact macroscopic flow equations for the case that the number of stored patterns is finite are derived, and the stability of attractors is evaluated. It is shown that the hysteretic property improves the robustness of both memory patterns and mixed states against temperature but does not affect the qualitative structure of the phase diagram. In addition, it is numerically showed that as the number of stored patterns increases and the state of the system becomes frustrated, the relaxation time becomes very long. In contrast to the results of the previous studies, however, memory capacity was not improved by the hysteretic property.

Image-based user profiling of frequent and regular venue categories

The availability of mobile access has shifted social media use. With that phenomenon, what users shared on social media and where they visited is naturally an excellent resource to learn their visiting behavior. Knowing visit behaviors would help market survey and customer relationship management, e.g., sending customers coupons of the businesses that they visit frequently. Most prior studies leverage meta-data e.g., check-in locations to profile visiting behavior but neglect important information from user-contributed content, e.g., images. This work addresses a novel use of image content for predicting the user visit behavior, i.e., the frequent and regular business venue categories that the content owner would visit. To collect training data, we propose a strategy to use geo-metadata associated with images for deriving the labels of an image owners visit behavior. Moreover, we model a users sequential images by using an end-to-end learning framework to reduce the optimization loss. That helps improve the prediction accuracy against the baseline as demonstrated in our experiments. The prediction is completely based on image content that is more available in social media than geo-metadata, and thus allows coverage in profiling a wider set of users.