مرکز راه‌کارهای اطلاعاتی هوشمند
Intelligent Information Solutions Center

Sharif University of Technology
فا | en
Homepage   >   R & D

R & D

IIS Center has always placed cutting edge R&D at the forefront of its activities, emphasizing the use of modern research findings in enabling technological solutions for its customers. To this end IIS Center's main research activities can be summed up into the following:

  • Knowledge creation in scientific & technical areas so as to expand products, processes and services
  • Collaboration with research labs and academic groups and presentation of scientific papers & articles in leading national and international journals and conferences
  • Providing research guidance to graduate students on defining and presenting theses centered on various industries
  • Creating collaboration platforms and venues for researchers so that they can focus on providing real practical value as part of their academic research

Also keeping in mind the fact that IIS Center is part of the Sharif University of Technology family and Sharif's role as Iran's & the region’s most prominent technical University we have created tried and tested processes so as to be able to commercialize new ideas and ventures with the help of the great talent at our disposal.

 



​Prediction and Multi-Scale Simulation of Cancer Cell Behavior​

Mohammad Hossein Zangooie   -   Feb 2016

Cancer involves the abnormal growth and spread of tissues within a body. Understanding the dynamics of cancer growth is one of the great challenges of modern science. Simulation of cancer growth in order to further elucidate this phenomenon and better diagnosis of cancer, is an important biological problem.

Little research has been done on the computer simulations of cancer growth which include differential equation, cellular automata, game theory and recently multi agent.

Current computational cancer modeling approaches can be divided into three categories: discrete, continuum, and hybrid. In this research we propose three scale simulation of cancer (macroscopic, mesoscopic and microscopic) to investigate it. To precise our simulation we consider it as a hybrid. In other words in each simulation scale combination of discrete and continuum approaches would be used.

Each cell considers as an agent at mesoscopic scale. This agent model introduces several new features such as angiogenesis. Our objective is to simulate the interaction between them at this scale. Also the cells biological function would be modeled by endowing each agent with a state.

At microscopic scale we need to simulate three networks (metabolism, signaling pathway and gene regulatory). By using differential equation we try to simulate inside and outside of cell at both microscopic and macroscopic scale.

Since the cancer simulation run time is too long, we want to extract features from a simulated model to predict the behavior of cancer.  We propose a learning method to use regression instead of classification. The proposed method was evaluated with several datasets and achieved acceptable accuracy. 

Optimizing Best Response Games on Social Networks

MohammadAmin Fazli   -   Jan 2016

Controlling dynamical systems is an important area of research, having many applications in related practical fields. One class of dynamical systems which have obtained ever-growing importance in recent years are networked dynamical systems, in which dynamic behavior is observed on nodes interacting in a networked setting. This is mainly due to the rise of complex networks, e.g. social networks, as an important and powerful method of capturing dynamic behavior that exists in different systems. One important aspect of such systems is that in many scenarios, actors that participate in them are selfish and act rationally based on their own self interest. Therefore guiding such systems to desirable states which may not be in the direct interest of the actors is a challenging and complex task.

To model the interactions of the actors in this context, game theory is used. Whilst classic game theory deals with static systems where the goal is to reach an equilibrium, an important subbranch of this area of research focuses more on the dynamics of reaching this equilibrium state. Therefore best response dynamics is used where in each round one actor, here referred to as a player, choses a best response keeping in mind his own utilization function and the current state of the network, i.e. each node's local state and also the state of edges between nodes.

In this thesis, we propose different methods based on combinatorial optimization to control the best response dynamics that ensues network games. In general, our target control problem is defined as an optimization problem which tries to find best values for the overall global parameters of the dynamical system which maximize a global objective function that calculates the desirability of the networks' state for us. In this thesis, different control parameters are considered, including 1) the initial setting of the dynamical system, 2) the underlying network, 3) the order in which players can take turns, 4) the cost parameters that can be used to penalize a player for some chosen action and 5) the set of players who are permitted to play. Therefore we can use the optimal values of these control parameters so that the interaction of the selfish players reaches a final equilibrium that is acceptable to us.

Using social network patterns to improve evolutionary algorithms

Behin Molaie   -   Aug 2015

Evolutionary algorithms are an important class of algorithms in the software field. Today evolutionary algorithms are used in many applications such as Artificial Art, Automated Design, Bio-informatics, Communications, etc.
Many articles has been published in this field. The Evaluation of this algorithms have two approaches, the quality of result and the runtime of the algorithm. In this thesis we proposed a new use of Social Networks in Genetic Algorithm, and we designed a parallel model of it on CUDA GPUs.
The main idea was that we seen a good content spread in the social networks very easy and fast. We used this idea to improve genetic algorithm by creating a social network between people of a generation. We used the most used benchmarks to evaluate our idea.
The second idea was to implement our idea on a massive parallel architecture and we implemented it on CUDA GPU architecture.

Analysis of Equilibrium Points in Economic Games

Pedram Beheshti   -   Jan 2015

Nowadays economical issuses are greatly taken into consideration. In this project we aim to study a problem of this area, which focuses on the subject of location allocation and pricing strategy for a good supplier. Solution evaluation is done by taking costumer demand into account.

This problem has a sequential characteristic. That is players do not take their decisions atthe same time. Suppliers should first decide about where and how they plan to provide their goods. on the other hand almost every economical problem have multiple players who play with or against each other. Therefore modeling this problem using Stackelberg leadership model is a natural choice.

Since this problem is in the NP-hard class, we aim to propose a heuristic solution. The proposed method divides this problem into two steps of location allocations and decideing on pricing policy. Simulation results show that using the sample, greedy location allocation method in conjunction with the proposed pricing strategy would provide desirable results.

Toward a Method for the Management of Dynamic Changes in Software Product Line Based on Change Patterns

Alireza Mazloumi   -   Sep 2014


With the advent of software product lines as a practical approach in software engineering, research on this area has increased exponentially. In Software product lines a platform is presented onto which a product family can be created with ease. With the help of this platform each software in the product family can be produced with small modifications. Using this approach, cost and time of producing software decreases whilst at the same time increasing the overall software quality.

Changes in standards, instability in business needs and variable requirements of various stakeholders will always cause evolution in a software product line. These change areas cannot always be predicted by variation points defined in the product line architecture, to the extent that it is possible for these changes to happen during the execution time of the product line. Management and maintenance of dynamic changes is an open challenge in the field of software product line research. Solving this challenge will have tremendous impact on the costs and performance associated with software product lines.

In this thesis we will propose a process in which management of dynamic changes will be handled. This process will act as an umbrella activity beside the product line execution methodology and will be responsible for the management of changes in the execution process. The main foundation for this process is the change execution cycle. This part of the process searches for and finds the best solutions amongst change patterns contained in the solutions repository , which has been created beforehand. At the end of the cycle the repository of patterns and solutions will be updated; thus in an iterative process the proposed method can learn and refine itself.

For investigating the performance of our proposed method, we demonstrate it on a real enterprise software system. After that, for the evaluation of this method, we took a survey of all relevant stakeholders of the enterprise system. The results showed that our proposed method achieves substantial gains in the process of change management.

Improving Network Structure and Behavior via Redundancy Injection with a Limited Budget

Mohsen Salari   -   Jan 2014

Improving the network structure is one of the most impostant challanges in network research. Improving network structure is performed in order to achieve goals such as refining routing and robustness.

Having an eye on the structure of a real packet switched communication network as an example, we have tried to improve its performance in those regards by injecting a small number if edges to the network in this project. To reach this goal, we have analyzied the current approaches in approximation algorithms in theoretical computer science, methods currently used in analyzing complex networks and heuristics that are based on network structure. Comparing the advantages and disadvantages of each approach, we have proposed a suboptimal strategy for edge injection, and using real network data verified the superiority of this method over the others.

Distributed Data Mining in Peer-to-Peer Systems

Hoda Mashayekhi   -   Jul 2013

peer-to-peer (P2P) computing is a popular distributed computing paradigm for many applications which involve exchange of information among a large number of peers. In such applications, large amount of data is distributed among multiple dispersed sources. Therefore, data analysis is challenging due to processing, storage and transmiSSiOn costs. Moreover, the data rarely remains static and frequent data changes, quickly out date previously extracted data mining models.

Distributed data mining deals with the problem of data analysis in environments with distributed data and computing resources. In this dissertation, we explore distributed data mining in different structures of P2P systems. In structured P2P systems, L-overlay is proposed for indexing data, and processing complex queries in P2P systems. The overlay is later used for K-nearest neighbor and Naïve bayes classification.

In unstructured P2P systems, gossiping proves to be an effective yet simple communication mean, which can also adapt to dynamics in the system. This communication paradigm enabled us to devise GoSCAN, a decentralized density-based clustering method which is adaptive to churn. The model is further extended to the novel decentralized algorithm GDCluster, which, to the best of our knowledge, is the first truly decentralized and adaptable clustering method applicable for different clustering algorithms.

The proposed methods enjoy scalability and incremental adaptation in presence of dynamics. Analysis ofthe algorithms and extended simulation results, show the robustness, effectiveness and scalability of the proposed methods under static and dynamic settings, with different data assignment strategies. Also different state-ofthe-aft methods such as SSW and LSP2P are employed for comparison purposes.

Evaluation of Data Models for the Purpose of Storage and Retrieval of Data and Knowledge Contained in Social Networks

Moslem Habibi   -   Sep 2012

Online Social Networks have become indispensable platforms to model social interactions and for the retrieval of knowledge based on economical, political, social and ... criteria. The importance of social networks for such uses stems from that vast amount of social data that it gathers for use by researchers. Aside from it's impact on the field of social network analysis, the ever-growing use of social media by users has added even more importance to social network data as it contains a history of the users virtual life.

Unfortunately research on proposals to support the management of this vast amount of data, wether for the conceptual modeling of social data or to provide a uniform modeling language to represent such data, has been highly lacking until recent times.

In this thesis we try yo provide a view into data model support for social networks, For this purpose we first explain extensions to well known data models like the relational model that are aimed at extending it's capabilities in regards to social network data support. Then new data models specifically aimed at social networks are introduced and then, by modeling a real world sample, compared to each other. In the end proposal for improvements to the models are considered.

On a Bounded Budget Network Creation Game

MohammadAmin Fazli   -   Mar 2011

In this thesis, we consider a network creation game in which, each player (vertex) has a limited budget to establish links to other players. In our model, each link has a unit cost and each agent tries to minimize its cost which is its local diameter or its total distance to other players in the (undirected) underlying graph of the created network. Two variants of the game are studied: in the MAX version, the cost incurred to a vertex is the maximum distance between that vertex and other vertices, and in the SUM version, the cost incurred to a vertex is the sum of distances between that vertex and other vertices. We prove that in both versions pure Nash equilibria exist, but the problem of finding the best response of a vertex is NP-hard.

Next, we study the maximum possible diameter of an equilibrium graph with n vertices in various cases. For infinite numbers of n, we construct an equilibrium tree with diameter θ(n) in the MAX version. Also, we prove that the diameter of any equilibrium tree is O(log(n)) in the SUM version and this bound is tight. When all vertices have unit budgets (i.e.~can establish link to just one vertex), the diameter in both versions is O(1). We give an example of equilibrium graph in MAX version, such that all vertices have positive budgets and yet the diameter is as large as Ω(√log(n)). This interesting result shows that the diameter does not decrease necessarily and may increase as the budgets are increased. For the SUM version, we prove that every equilibrium graph has diameter O(√n) when all vertices have positive budgets. Moreover, if the budget of every players is at least k, then every equilibrium graph with diameter more than three is k-connected.

A Code Generation Template for the Transformation of UML Diagrams to the Symfony Framework

Moslem Habibi   -   Sep 2010

The ever growing use of model-driven process in software engineering has made possible the automatic transformation of models used in such processes to source code. However little work has been done on the transformation of the ever popular UML diagrams to web based application. In this thesis such a generator for the aforementioned task is presented, using Symfony, a PHP based framework, as our source language.