Research / Projects

(Postgraduate) Computational Neuroscience

Neural Network Modelling of the Primate Ventral Visual Pathway

2012 - 
supervisor: Dr. Simon Stringer
Eguchi
Over successive stages, the primate ventral visual pathway develops neurons that respond to particular objects or faces independently of their position, size or orientation. The ventral visual pathway is thus thought to be responsible for transform-invariant visual object and face recognition in the brain. However, it remains a difficult challenge to understand exactly how these neurons develop their response properties during learning. The learning processes will depend on how the neurons interact with each other through successive layers of the ventral visual pathway as they are driven by rich visual input from natural scenes. This can only be investigated through computer simulations that accurately model the behaviour of individual neurons, how these neurons are linked together in the brain, how the synaptic connections between cells are modified during learning, and the statistical properties of the visual input from the sensory environment.

Publication / Presentation

  • Akihiro Eguchi, Simon Stringer, Glyn Humphreys. "The visually-guided development of facial representations in the primate ventral visual pathway: a computer modeling study," Psychological Review (IF 7.58), Volume 123, Issue 6, Nov, 2016. 
  • Courtney Spoerer*, Akihiro Eguchi*, Simon Stringer. "A computational exploration of complementary learning mechanisms in the primate ventral visual pathway," Vision Research (IF 2.41), Volume 119, pp 16–28, Jan 2016, doi: http://doi.org/10.1016/j.visres.2015.12.008 [pdf]
  • Akihiro Eguchi, Bedeho Mender, Benjamin Evans, Glyn Humphreys, Simon Stringer. "Computational Modeling of the Neural Representation of Object Shape in the Primate Ventral Visual System," Frontiers in Computational Neuroscience (IF 2.2). Volume 9, Issue 100, 4th Aug., 2015, doi:10.3389/fncom.2015.00100 [pdf]
  • Akihiro Eguchi, Samuel Neymotin, Simon Stringer. "Color opponent receptive fields self-organize in a biophysical model of visual cortex via spike-timing dependent plasticity," Frontiers in Neural Circuits (IF 3.3), Volume 8, Issue 16, 2014, doi:10.3389/fncir.2014.00016. [pdf]

(Undergraduate) Computer Science Projects

Smart Housekeeper Robot with Android

2011 - 2012
supervisor: Dr. Craig W. Thompson
Eguchi, Nguyen, Gouveia
In this project we are aiming to establish an architecture to remotely control a mobile robot that can collect information of all devices around. The information can be used to retrieve the API of associated devices from a database on a server so that people can control those devices via the mobile robot over the internet. The mobile robot is embedded with an Android phone so that it can be easily connected to the internet to stream a video image to navigate the robot and control the surrounding objects.

Publication / Presentation

...more details

Object recognition and activity learning using Kinect

2011
independent project
Eguchi
This project explores a new approach to computational object recognition by borrowing an idea from child language acquisition studies in developmental psychology.  Whereas previous image recognition research used shape to recognize and label a target object, the model proposed in this thesis also uses the function of the object resulting in a more accurate recognition.  This thesis makes use of new gaming technology, Microsoft’s Kinect, in implementing the proposed new object recognition model.  A demonstration of the model developed in this project properly infers different names for similarly shaped objects and the same name for differently shaped objects.

Publication

Ontology in a virtual world

2010
supervisor: Dr. Craig W. Thompson
Eguchi
Developing an ontology service by organizing the data of objects consisting of explicit identity and type information collected from a virtual world by Josh Eno (Virtual and Real-World Ontology Services). This information would construct a semantic network ontology graph which can have many uses from keeping track of assets like wheelchairs to controlling or querying objects in the real world.

Autonomous Floor Mapping Robot project

2010 - 2011
independent project from any faculty
Eguchi, Nguyen, Hooten
The first focus of this research was the development of an autonomous floor mapping robot. In order to reduce the cost of building prototypes to test the program, we used the Simbad 3D simulator. To test in a more realistic environment, we established a way to control objects in a virtual world Second Life. Then, for the hardware part, we built a low cost robot with inexpensive, but accurate, Sharp IR sensors with a regular optical mouse.

Publications

Knowledge Representation in Smart Objects

2010 - 2011
supervisor: Dr. Craig W. Thompson
Eguchi
Identifying protocols that smart objects will need to follow and how to use today’s 3D virtual worlds to better design and understand those protocols for tomorrow’s smart world. I developed the architecture in a virtual world that exclusively uses cloud computing so that any object can be seen as a smart object without embedded intelligence.

Publications

  • Akihiro Eguchi, Craig W Thompson. “Towards a Semantic World: Smart Objects in a Virtual World,” International Journal of Computer Information Systems and Industrial Management, Volume 3, 2011. pp. 905-911. [View Download]
  • Akihiro Eguchi, Craig W Thompson. “Smart Objects in a Virtual World,” Inquiry Journal of Undergraduate Research, University of Arkansas, Volume 11, 2010. pp. 3-8.
  • Akihiro Eguchi, Craig W Thompson. “Towards a Semantic World: Smart Objects in a Virtual World,” Web Virtual Reality and Three-Dimensional Worlds Workshop (IADIS WEB3DW2010), Freiburg, Germany, July 26-31, 2010. pp.488-493.
  • Akihiro Eguchi. "Smart Objects in a Virtual World," Poster presentation. Cyber Infrastructure Days Conference, May 16-17, 2010. http://hpc.uark.edu/cidays/
  • Akihiro Eguchi, Craig W Thompson. “Smart Objects in a Virtual World,” Conference on Applied Research in Information Technology, Acxiom Laboratory for Applied Research, Conway AR, April 9, 2010. 
  • Akihiro Eguchi, Craig W Thompson. “Smart Objects in a Virtual World,” X10 Workshop on Extensible Virtual Worlds, venue: Second Life, March 29 - April 2, 2010. http://vw.ddns.uark.edu/X10


Minority Game project

2010
Independent project from any faculty
Eguchi, Nguyen
Developed an extended simulation model of the minority game, which is a variant of the El Farol Bar proposed by professors from the U. of Fribourg. In this simulation, normal agent (using Genetic Algorithm), super agent (using Statistical Analysis), group agent (group of normal agent) and human agent (multiple users connected through network using server/client architecture) compete in zero-sum game.

Publications


(Undergraduate) Psychology Projects

Children's Language Acquisition

2011 - 2012
supervisor: Dr. Douglas A. Behrend
Eguchi
Children are known to use various learning biases to efficiently develop their language skills. Prior studies have confirmed that young children keep track of reliability histories of possible teachers in order to selectively learn words. Furthermore, it has been shown that they are less likely to learn from foreign language speakers or foreign names of objects. Recent studies even indicated that children seem to change their patterns of learning just by hearing the cultural background of the target objects. The purpose of this thesis is to investigate the cultural bias in word learning among 2- to 3-year-olds. We hypothesized that children would be less likely to learn object names and functions when they were told those objects and/or the teacher were from another country. The result of the experiment showed no evidence to support this hypothesis; however, we found that participants performed very poorly in the novel word learning task, yet very well on the novel function learning task. Possible reasons for these findings are discussed.

Papers


Self-perception study: Cultural Differences in Identity Perspectives

2011
supervisor: Dr. Scott Eidelman
Eguchi, Bohannan
Self-categorization theory explains that our process of self-relevant information is more context-dependent rather than a reflection of our self-schema. This theory proposes two types of identities (a personal identity and a social identity), and a salience of the identity can void the effect of self-schema. Those two identities are thought to show antagonism, but because the levels of both identities vary depending on the population and on the degree of self-awareness, the hypothesis is questionable. The current study focuses on different characteristics between an individualistic culture (the U.S.) and a collectivistic culture (Japan) in order to test the validity of antagonism. The experiment is conducted via the Internet using a carefully designed computer mediated communication (CMC) to deal with the minority ethnicity effect and a false identity construction. The result showed that in a self-focused condition with a web-cam, American participants decreased the social identity; conversely, Japanese participants increased the social identity. This contrast successfully showed the cultural differences in identity perspectives. However, unlike many past studies with a mirror, both American and Japanese participants decreased the personal identity in the condition with a web-cam.

Presentation / report

  • Akihiro Eguchi, Emily Bohannan. "Who do you see in a mirror? cultural differences in identity perspectives," Poster presentation. Advanced Research Poster Session, University of Arkansas, Dec 06, 2011. [View Download]
  • Akihiro Eguchi. "Who do you see in a mirror? cultural differences in identity perspectives," Report submitted in advanced Research course, University of Arkansas, Dec 14, 2011 [View Download]