Month: February 2008

Neural Network: implementing backpropagation

In this post i report a simple implementation of a Backpropagation Neural Network.
As for the following examples i will post the goal is character recognition on a grid of pixels. 
Noise is introduce to evaluate robustness of the approach.
  
Here we will present both implementation and performance eval- 
uation of a multilayer artificial neural network, exploiting backpropagation for characters recognition.
 
This documentation provide full description of the approach and the results.
This is the source code for the neural network

Iphone 1.1.2 OTB Full Software Unlock: finally!!

It seems that George Hots made it to software unlock Iphone 1.1.2 OTB 4.6 and chances are that this will work for 1.1.3 as well. 
 
here his blog:  

http://iphonejtag.blogspot.com/2008/02/11246unlock-good-enough-for-prize.html

 

 I will personally wait a little, since the guy upgrade a baseband 3.9 to 4.6 and then found the exploit. I want to make sure that a pure native 4.6 works. As soon as the solution prove to be stable i’ll go through it and let yo know. But it seems that the dawn is close… 

 

and as often iClarified make hard things easy with a fool-proof tutorial:

http://iclarified.com/entry/index.php?enid=649 

 

Neural Network: implementing Madaline

In this post i report a simple implementation of a Madaline Neural Network.
As for the following examples i will post the goal is character recognition on a grid of pixels. 
Noise is introduce to evaluate robustness of the approach.
  
Here we will present both implementation and performance eval- 
uation of a multilayer artificial neural network, based the Adaline 
neurons (Madaline) for characters recognition.
 
This documentation provide full description of the approach and the results.
This is the source code for the neural network

Wireless Sensor Networks: the TinyLime experience!

In the the past i spent some time working on Wireless Sensor Networks. The most complete result of this research has been the design and development of TinyLIME, in collaboration with GianPietro Picco and Amy Murphy, the original designer and developer of LIME.

TinyLIME is a middleware for wireless sensor networks (WSN) that departs from the traditional WSN setting where sensor data is collected by a central monitoring station, and enables instead multiple mobile monitoring stations to access the sensors in their proximity and share the collected data through wireless links. This intrinsically context-aware setting is demanded by applications where the sensors are sparse and possibly isolated, and where on-site, location-dependent data collection is required. An extension of LIME, TinyLIME makes sensor data available through a tuple space interface, providing the illusion of shared memory between applications and sensors. Data aggregation capabilities and a power-savvy architecture complete the middleware features.

 Please refer to the official webpage for more details and to download the system. 

AI simple programs: MiniMax for game playing

Another interesting topic in AI is game playing, the MiniMax algorithm is been presented,
together with the optimized version known as Alpha-Beta pruning. Here i show an
implementation and result evaluation for both algorithm on a simplified version of the
well known Othello Game (also known as Reversi or Attaxx). 

 This documentation provide a complete description of the steps and  
here you can find the source code for the optimized version of the program. 

AI simple programs: heuristic search

Informed search algorithm are usefull when dealing with large state 
space, and when ”good” heuristics can be find for the problem. Here we 
present two simple an implementation of the well known algorithms: A*. 
To show this two simple game, the 8 puzzle game and the dating game has 
been used, in the previous we tested 4 different heuristics while the second 
has been used to compare simple depth first, A* and IDA*. 
 
 This documentation provide a complete description of the steps and  

here you can find the source code for the optimized version of the program. 

AI simple programs: searching

 
Searching algorithm are commonly exploited in the Artificial Intelligence field, 
here I present two simple implementations of the well known 
algorithms: depth first, breadth first and depth limited for tree search. To 
show this two simple game, the 8 puzzle game and a simple grid world 
has been used.
 
This documentation provide a complete description of the steps and  
here you can find the source code for the optimized version of the program. 
 

DrugSearch: Neural Network Information Retrieval

 

In this post is summarize the work i’ve done as Final Project for CS580 Neural Network in Chicago for Professor Graupe. This work has been later extended together with professor Clement Yu and published at CIKM 2005.  General purpose search engines have provided the users with simple way of querying massive sources of information efficiently, but for domain specific problems better solutions can be developed. We present an application of the Lamstar Neural Network for domain specific web searches. The problem itself is really general and is about relating different topics in a domain specific web search, we present here an instance of the problem relating drug names and side effects. The system, working on top of existing search engines, will filter the results to improve the quality of the search via a properly trained Lamstar neural network, providing a list of links to the suitable pages while directly showing the most relevant portion of each page to the user. 

 

Here is a presentation of the work: drugsearch.pdf

Software Refactoring

The following presentation is a joint work with Giorgio Orsi on Software Refactoring, is one of the intermediate product of the course Argomenti Avanzati in Ingegneria del Software (Advanced Topic of Software Engineering). It is just a summa of the existing approaches, and some of our opinions on how a disciplined Refactoring can help to reduce the gap between Research and Didactic, by help in the integration of student activities (Projects and Thesis) in the main stream of professors research. Refactoring can indeed fill up the gap from the so called student-ware and the final software.

refactoring.pdf 

NSPACE = CO-NSPACE

For the course ADVANCED TOPICS ON ANALYSIS AND DESIGN OF CRITICAL SYSTEMS    i attended in Politecnico di Milano in 2007, I had to study and present the Neil Immerman proof of NSPACE = CO-NSPACE. This is the small presentation i prepared with Simone Campanoni. We present the two available proofs Szelepcsènyi’s and Immerman’s. I hope this will make someone life easier.

 nspace.pdf