Search Results

Search found 22716 results on 909 pages for 'network architecture'.

facebook twitter digg google delicious technorati stumbleupon myspace wordpress linkedin gmail igoogle windows live tumblr viadeo yahoo buzz yahoo mail yahoo bookmarks favorites email print

Page 305/909 | < Previous Page | 301 302 303 304 305 306 307 308 309 310 311 312  | Next Page >

  • Neural Networks in C# using NeuronDotNet

    - by kingrichard2005
    Hello, I'm testing the NeuronDotNet library for a class assignment using C#. I have a very simple console application that I'm using to test some of the code snippets provided in the manual fro the library, the goal of the assignment is to teach the program how to distinguish between random points in a square which may or may not be within a circle that is also inside the square. So basically, which points inside the square are also inside the circle. Here is what I have so far: namespace _469_A7 { class Program { static void Main(string[] args) { //Initlaize the backpropogation network LinearLayer inputLayer = new LinearLayer(2); SigmoidLayer hiddenLayer = new SigmoidLayer(8); SigmoidLayer outputLayer = new SigmoidLayer(2); new BackpropagationConnector(inputLayer, hiddenLayer); new BackpropagationConnector(hiddenLayer, outputLayer); BackpropagationNetwork network = new BackpropagationNetwork(inputLayer, outputLayer); //Generate a training set for the ANN TrainingSet trainingSet = new TrainingSet(2, 2); //TEST: Generate random set of points and add to training set, //for testing purposes start with 10 samples; Point p; Program program = new Program(); //Used to access randdouble function Random rand = new Random(); for(int i = 0; i < 10; i++) { //These points will be within the circle radius Type A if(rand.NextDouble() > 0.5) { p = new Point(rand.NextDouble(), rand.NextDouble()); trainingSet.Add(new TrainingSample(new double[2] { p.getX(), p.getY() }, new double[2] { 1, 0 })); continue; } //These points will either be on the border or outside the circle Type B p = new Point(program.randdouble(1.0, 4.0), program.randdouble(1.0, 4.0)); trainingSet.Add(new TrainingSample(new double[2] { p.getX(), p.getY() }, new double[2] { 0, 1 })); } //Start network learning network.Learn(trainingSet, 100); //Stop network learning //network.StopLearning(); } //generates a psuedo-random double between min and max public double randdouble(double min, double max) { Random rand = new Random(); if (min > max) { return rand.NextDouble() * (min - max) + max; } else { return rand.NextDouble() * (max - min) + min; } } } //Class defines a point in X/Y coordinates public class Point { private double X; private double Y; public Point(double xVal, double yVal) { this.X = xVal; this.Y = yVal; } public double getX() { return X; } public double getY() { return Y; } } } This is basically all that I need, the only question I have is how to handle output?? More specifically, I need to output the value of the "step size" and the momentum, although it would be nice to output other information as well. Anyone with experience using NeuronDotNet, your input is appreciated.

    Read the article

  • July 17th Live Webcast with Oracle's Tom Kyte

    - by jgelhaus
    Webcast: Oracle Maximum Availability Architecture Best Practices Date: Tuesday, July 17, 2012 Time: 10 a.m. PT/1 p.m. ET Update Your Knowledge with Oracle Expert Tom Kyte With Oracle’s Maximum Availability Architecture (MAA), organizations can minimize the cost and risk associated with downtime. Oracle’s MAA best practices extend beyond Oracle Database to span a broad range of products, including Oracle Exadata and Oracle Database Appliance. Join Oracle expert Tom Kyte for this interactive Webcast to learn how to: Protect your systems from planned and unplanned downtime Achieve the highest quality of service at the lowest cost Eliminate idle redundancy in the data center Register today and ask Tom your questions around availability best practices.

    Read the article

  • Open Source/Free Social Networking SDK

    - by Nix
    I am currently gathering some technology requirements for a site that will be a social network based. I don't want to re-invent the wheel so i am looking for some type of SDK or collection of tools that can provide me with a way of creating/managing a social network. I understand that no framework will probably fit my exact needs so I am also looking for a flexible/extensible framework. An example extension point would be allowing the user to provide sub networks, maybe a global network that could be sub classified as work and friends. Beyond that it would also be nice to somehow be able to import contacts from other networking sites (Facebook, Linked In, etc). My current technology suite will more than likely consist of the following: IIS 7.0 WCF Data Services SQL Server 2005/2008 ASP.NET front end. So my two questions are 1) C# Open Source Social Network SDK 2) C# Open Source Social Netowrk APIs (facebook, linked in, etc) If there is any more information you may need please let me know.

    Read the article

  • Optimizing collision engine bottleneck

    - by Vittorio Romeo
    Foreword: I'm aware that optimizing this bottleneck is not a necessity - the engine is already very fast. I, however, for fun and educational purposes, would love to find a way to make the engine even faster. I'm creating a general-purpose C++ 2D collision detection/response engine, with an emphasis on flexibility and speed. Here's a very basic diagram of its architecture: Basically, the main class is World, which owns (manages memory) of a ResolverBase*, a SpatialBase* and a vector<Body*>. SpatialBase is a pure virtual class which deals with broad-phase collision detection. ResolverBase is a pure virtual class which deals with collision resolution. The bodies communicate to the World::SpatialBase* with SpatialInfo objects, owned by the bodies themselves. There currenly is one spatial class: Grid : SpatialBase, which is a basic fixed 2D grid. It has it's own info class, GridInfo : SpatialInfo. Here's how its architecture looks: The Grid class owns a 2D array of Cell*. The Cell class contains two collection of (not owned) Body*: a vector<Body*> which contains all the bodies that are in the cell, and a map<int, vector<Body*>> which contains all the bodies that are in the cell, divided in groups. Bodies, in fact, have a groupId int that is used for collision groups. GridInfo objects also contain non-owning pointers to the cells the body is in. As I previously said, the engine is based on groups. Body::getGroups() returns a vector<int> of all the groups the body is part of. Body::getGroupsToCheck() returns a vector<int> of all the groups the body has to check collision against. Bodies can occupy more than a single cell. GridInfo always stores non-owning pointers to the occupied cells. After the bodies move, collision detection happens. We assume that all bodies are axis-aligned bounding boxes. How broad-phase collision detection works: Part 1: spatial info update For each Body body: Top-leftmost occupied cell and bottom-rightmost occupied cells are calculated. If they differ from the previous cells, body.gridInfo.cells is cleared, and filled with all the cells the body occupies (2D for loop from the top-leftmost cell to the bottom-rightmost cell). body is now guaranteed to know what cells it occupies. For a performance boost, it stores a pointer to every map<int, vector<Body*>> of every cell it occupies where the int is a group of body->getGroupsToCheck(). These pointers get stored in gridInfo->queries, which is simply a vector<map<int, vector<Body*>>*>. body is now guaranteed to have a pointer to every vector<Body*> of bodies of groups it needs to check collision against. These pointers are stored in gridInfo->queries. Part 2: actual collision checks For each Body body: body clears and fills a vector<Body*> bodiesToCheck, which contains all the bodies it needs to check against. Duplicates are avoided (bodies can belong to more than one group) by checking if bodiesToCheck already contains the body we're trying to add. const vector<Body*>& GridInfo::getBodiesToCheck() { bodiesToCheck.clear(); for(const auto& q : queries) for(const auto& b : *q) if(!contains(bodiesToCheck, b)) bodiesToCheck.push_back(b); return bodiesToCheck; } The GridInfo::getBodiesToCheck() method IS THE BOTTLENECK. The bodiesToCheck vector must be filled for every body update because bodies could have moved meanwhile. It also needs to prevent duplicate collision checks. The contains function simply checks if the vector already contains a body with std::find. Collision is checked and resolved for every body in bodiesToCheck. That's it. So, I've been trying to optimize this broad-phase collision detection for quite a while now. Every time I try something else than the current architecture/setup, something doesn't go as planned or I make assumption about the simulation that later are proven to be false. My question is: how can I optimize the broad-phase of my collision engine maintaining the grouped bodies approach? Is there some kind of magic C++ optimization that can be applied here? Can the architecture be redesigned in order to allow for more performance? Actual implementation: SSVSCollsion Body.h, Body.cpp World.h, World.cpp Grid.h, Grid.cpp Cell.h, Cell.cpp GridInfo.h, GridInfo.cpp

    Read the article

facebook twitter digg google delicious technorati stumbleupon myspace wordpress linkedin gmail igoogle windows live tumblr viadeo yahoo buzz yahoo mail yahoo bookmarks favorites email print

< Previous Page | 301 302 303 304 305 306 307 308 309 310 311 312  | Next Page >