to see the past projects scroll down.
Script-Draw Programming Language
I am designing a new programming language for creative coding. Instead of being a compiled language/API like Processing, OpenFrameworks and Cinder, it is intended to be a scripting language with dynamic typing where the user is provided with a REPL to be able to evaluate the results dynamically and interactively.
In the most basic usage, a short script is written to provide a drawing environment. Then user draws lines on the screen which passes a list of points. After that the list would be easily accessible through the scripting environment. The user can use the coordinates and order of points to create different forms.
The same process can be used to produce audio or general analog signals and can drive various external devices.
Beating Hypothesis for various time-scales in complex networks
Beating is a rather well-known concept but it has some interesting and not very much studied aspects. In this report, I have demonstrated the robustness of beating phenomena under various sorts of noise and disturbance.
The main idea of the report is that beating can be responsible for existence of different time-scales in complex networks. Here I have speculated some of the properties of time perception in the brain through a simple beating clock model.
Python API for Allen Connectivity Atlas
I am designing an easy to use API for Allen Connectivity Atlas. This API will give the developers the ability to fetch, search and visualize the connectivity data and underlying images. Furthermore, it is planned to let the user run basic data mining and machine learning tasks.
Projects (Outdated -.- )
Communicability in Complex Networks
In this project, I have discussed the properties of network communicability. Then I have proved a new bound for it which relates its value to the number of edges in the graph.
Next I have defined a new measure of communicability based on physical networks and exponential signal attenuation (Beer–Lambert law). The new measure does a slightly better job at recovering the topology of the network.
At the end, I have used the measures of communicability to analyze the brain connectivity network, before and eight weeks after stroke.
As a part of my MSc thesis, I designed a method to acquire an overcomplete dictionary of cortical evoked signals. In this method, we stimulate the transgenic mouse (ChR2) cortex at a grid of equispaced points. The grid is automatically generated based on the size and geometry of the craniotomy window and is registered at the Hind-Limb primary sensory cortex. The stimulations are done using a robotic system to guid the laser beam. In this project, I designed the experiment and programmed the robotic system to stimulate the animal's brain.
Calibrated Sensing for Sparse Recovery of Video
I have developed a strategy to tackle complexity/RIP dilemma when dealing with sparse recovery of shift invariant unions of subspaces.
In short, the idea is to calibrate the sensing matrix according to energy distribution along the signal. This way, the sensing matrix gets much smaller and also the restricted isometry property of it gets improved. The results are in preparation for publication.
Shift-Adaptive Sparse Recovery for Video
This is the more advanced case of the previous project where I tune the shifts in the sensing matrix iteratively (using simple binary search). This method adds to the complexity of the algorithm but improves the accuracy of the results significantly. The results are in preparation for publication.
Brownian Bridge Vectorization
I used two dimensional brownian bridges, morphological segmentation of the images and approximate solutions of travelling salesman problem to create these visualizations. Read the full report to see how !
Note that the method is easily implemented using the Script-Draw programming language.
Mohajerani MH *, Chan, AW *, Mohsenvand M, LeDue J, Liu R, McVea D, Boyd J, Reimer M, Wang YT, Murphy TH (2013). Nature Neuroscience 16;1426–1435 PMID: 23974708 (*equal co-first author) (Selected for Journal Cover)
Optical Flow Analysis of VSD signals
I used optical flow algorithms to measure velocity of cortical waves. Then I analyzed the resulting vector fields to find the cortical sinks and sources.
The results of my work was published in the journal of Nature Neuroscience. Look at the section above !
Rod Theory applied to Foldable Tents
I used the mathematics of rod theory to mathematically model the behavior of foldable tents.
The rod theory has recently gained popularity in the analysis of mechanics of DNA molecules. It has some applications in computer graphics as well.
Real-time Neural Field Simulator
This project is a real-time neural field simulator which uses simple Eulerian integration to solve the integro-differential equations.
Using this simulator, I was able to simulate different kinds of pattern formation.
Real-time Satellite attitude Control System
In this project I built a satellite attitude control system. I could change the attitude of the satellite, apply noises and distortions using a joystick and see the results as a virtual reality. All the project was done in Simulink.
Tumor growth simulation using Cellular Automata
For this project, I wrote a couple of algorithms to simulate tissue growth under various conditions.
The image shows a tumour simulation using my codes.
Seizure suppression using Optogenetics
This project was based on a straightforward idea for desynchronization and inhibition of seizure activity in epileptic patients using optogenetics.
The viral vector has two parts (T) for detection and marking of the epileptic area and (S) for infecting the area with NpHR. The area also expresses XFP so that surgeon can detect the area.
Compact Topology Affects Central Forces
In this letter I have discussed an interesting mathematical speculation about central forces on manifolds with compact dimensions. Here for the sake of simplicity, I have shown the effects for a 2D electrostatic- like interaction between two pointlike charged particles embedded on a cylindrical space. It turns out that the form of the interaction differs from the case of flat 2D plane. A very remarkable result is that for long distance interactions, the strength of the force doesn’t approach zero and instead, it tends to a constant relative to the inverse of the radius of the cylinder. By the way, for very large redii of the cylinder or for very small distances, the interaction asymptotically approaches the classical form. At the end, I have discussed some possible implications and suggestions for future works.
Unknotting Behavior of Knots: Combinatorial Unknotting Graph
This was an idea about combinatorial representation of unknotting process in knots.
I invented a simple polynomial representation for the unknotting process of knots based on the resulting graph structure after switching the crosses of the knot.
From Neural Networks to Social Networks
Here I have informally discussed a few ideas on modeling social force phenomena (as stated by Deb Roy) through a simplistic rate-based neural network model. My goal is to address the social F=MA law in a more rigorous way and point out some of its inefficiencies. In doing so, I have introduced a simple visualization method as a thinking tool for Milgram (crowd gaze) experiment. This visualization is capable of demonstrating some of the complex behaviors of social networks such as competition bias and attention inversion. At the end, I have presented a modified model that takes such behaviors into account
Ontology Generation From Unstructured Data
Fundamental Laws of Story Telling
Beauty and the Code
This project is an attempt towards a better understanding of computational cognitive science of beauty/interestingness per- ception (BIP). Here we will try to examine and improve a theory called Coherence Progress (or Compression Progress) by Ju ̈rgen Schmidhuber which puts forward an information theoretic framework for BIP and has been used in optimiza- tion of intelligent agents (Schmidhuber (2008, 2010); Schaul, Pape, Glasmachers, Graziano, and Schmidhuber (2011)). In order to come up with a computationally expressible model, we took mathematical formulas as the generative algorithms of 2-dimensional plots and examined changes in beauty/inter- estingness rating (BIR) of human subjects with respect to per- ceived complexity of the formulas. First we presented tens of formulas to more than a 100 human subject and modeled their formula complexity perception (FCP) using bayesian and evo- lutionary approaches. Then we used the obtained models to generate 2d plots of functions with a broad range of FCP. Next we showed the plots with two different kinds of formulas (sim- ple and complex) to more than 200 human subject and gathered their BIR before and after seeing the formulas. By analyzing the results, we could confirm two hypotheses: (i) Rating of In- terestingness of a plot increases after showing a mathematical formula that is used to generate the plot and (ii) Rating of In- terestingness of a plot after seeing a simple formula is higher than its rating after seeing a complex formula. Both results are in agreement with Schmidhuber’s theory. At the end, we mod- eled the relationship of change in BIR with FCP and speculated an improved version of Schmidhuber’s work.
Two Projects to be added Here :
2- Motif-Orbit Embedding