computer science

To model the action potential of granule cell I employed the realistic model called Hodgkin – Huxley model and for modeling dendritic trees of granule cell I used multicompartmental model of neuron. In this presentation I will give my reasons about why I used theses models and at the end I will talk about some of my results.

Hamza will be presenting what he has been working on last few months. If you are interested to see some 3D active shape models (ASM) of lumbar vertebrae please come along.

Deep learning is increasingly popular due to its impressive performance. In this presentation, I am going to cover some basic concepts and the architectures when it is used in computer vision. Later, we can discuss how it works and its relation to other machine learning tools.

Last week I attended a talk by Daniel MacArthur, as a part of the Australian Neurogenetics conference. He has been leading the Human Knock-out Project, which aims to investivate the loss-of-function variants that healthy humans carry on their (our) genomes. I'll give a 30 minutes presentation on why an average healthy human has about 20 (normally functional) genes knocked-out. And why in some particular cases, a non-functioning gene can lower our cholesterol, increase our endurance, or protect us against HIV.

This talk is about a new project that I have started working on with Ali, Brendan and Martin. We are attempting to build a computational model for object recognition from images with internal representation identifying prototypical object types. The aim is to identify a particular instance of a type by measuring how that instance varies from the prototypical type. This is motivated by Ali's ideas on the connection between language and sensory-motor aspects of cognition...about which there will be very little in this talk (especially since Ali's away). Instead, I will talk about our efforts to build this model.

The aim of the project is to implement an unsupervised visually guided reaching routine on a robotic system. This system is composed by a four degrees of freedom robotic arm, two cameras and a 3D printed grabber with touch sensors on its interior and exterior surfaces. The idea is to develop an adaptive algorithm, which one can be transferable on others robotic arm, even if they don’t have the same characteristics (numbers of motors, size of pieces…). The robot will be able, after an unsupervised training process to reach and grasp objects. An interesting way to develop this kind of algorithm is to analyze how the human brain learn to plan trajectories, to reach and grasp movements.  In this presentation I will do an overview of the project.

In this talk, I will give a brief explanation of Honeybee segmentation process and the related genes. Then, I will show how we apply ODEs to model the genes and predict a gene network which forms early segmentation in Honeybee embryos. I also discuss the finding results.

In recent work we have observed that a new spike-timing-dependent model of metaplasticity has the effect of significantly enhancing the size of polychronous neural groups (PNGs). The relevance of polychronicity in the information processing of biological networks has been the primary focus of my thesis, with particular emphasis on the potential role of PNGs in representation and memory. In this talk I will describe the observed effect in more detail and attempt to provide an explanation. An understanding of the underlying mechanism for this enhancement is likely to shed light on the principles of PNG formation and perhaps therefore also on the processes of memory formation and storage.

This talk will explore the PoseCut algorithm for segmentation and pose estimation. Given a model for foreground and background and a simple model of the target the algorithm automatically finds the optimal segmentation and pose in each frame using graph cuts. I will also detail possible extensions to the algorithm.

I have investigated the synaptic plasticity of the granule cell based on the real experimental studies in two different protocols. In the first protocol spontaneous activity was applied to the both Medial and lateral pathways of the granule cell. However, HFS (high frequency stimulation) was applied just to the medial pathway. Therefore synaptic weight increased in the medial pathway while, synaptic weight decreased in the lateral pathway. In the second protocol the same as the first one we applied spontaneous activity to the both medial and lateral pathways while during medial HFS we set the lateral activity off. Synaptic weight increased in the medial pathway, however in the lateral pathway synaptic weight did not change.

Going back in time, this talk is about different kinds of industry/research projects Claudia was involved with as a software developer before she started the PhD. Related fields are CAD, augmented reality, virtual reality and interaction design.

Hamza will be presenting his work within last few months on the 3D construction of vertebrae, some validation process and 3D alignment of cloud points.

Glenn will discuss experimental results from his PhD work on representing logic in neural networks.

Xiping will discuss some approaches which can be used to speedup the nearest neighbour searching process. Specifically, he will focus on tree based and hashing based algorithms and show how they are used, and then discuss his recent work in this field.

Maria will give a brief introduction to particle filters, and then discuss her recent work using particle filters for face tracking. This work is part of her PhD research on the watching window project

I'll talk about some basic properties of proteins using a snake puzzle cube as a model. This was part of my Masters some years ago.

Given a video sequence, which captures a static scene/rigid object from various viewpoints, we want to make a 3D model of what the camera is looking at. In this presentation I will talk about the basics of 3D reconstruction using the structure from motion (SfM) approach: how it is supposed to work, why it’s harder than I initially thought, and our current results.

I am going to show a video which tells about the evolution of animal eyes, from a simple eye which can only recognize dark and light to a complex eyes. It also tells about the differences between predator and prey eyes.

I'll be talking about a neural network model of the storage of episode representations in working memory (WM) that Ali and I have been developing. Our key idea is that episodes are encoded in WM as prepared sensorimotor routines: i.e. as prepared sequences of attentional and motor operations. Our network reproduces several experimental findings about the representation of prepared sequences in prefrontal cortex. Interpreted as a model of WM episode representations, it has useful applications in an account of long-term memory for episodes and in accounts of sentence processing.

How do you approach a research or programming project? What do you do when planning? When do you start writing code? Writing reports? How do you organise code and data and experiments and documents and stuff? This week I'll be leading a discussion around these and other issues. Come along and share what you do, and learn from what others do.