Search Results

Search found 1366 results on 55 pages for 'artificial intelligence'.

Page 15/55 | < Previous Page | 11 12 13 14 15 16 17 18 19 20 21 22  | Next Page >

• Genetic Programming Online Learning

  - by Lirik

  Has anybody seen a GP implemented with online learning rather than the standard offline learning? I've done some stuff with genetic programs and I simply can't figure out what would be a good way to make the learning process online. Please let me know if you have any ideas, seen any implementations, or have any references that I can look at.

  Read the article

• Rush Hour - Solving the game

  - by Rubys

  Rush Hour if you're not familiar with it, the game consists of a collection of cars of varying sizes, set either horizontally or vertically, on a NxM grid that has a single exit. Each car can move forward/backward in the directions it's set in, as long as another car is not blocking it. You can never change the direction of a car. There is one special car, usually it's the red one. It's set in the same row that the exit is in, and the objective of the game is to find a series of moves (a move - moving a car N steps back or forward) that will allow the red car to drive out of the maze. I've been trying to think how to solve this problem computationally, and I can really not think of any good solution. I came up with a few: Backtracking. This is pretty simple - Recursion and some more recursion until you find the answer. However, each car can be moved a few different ways, and in each game state a few cars can be moved, and the resulting game tree will be HUGE. Some sort of constraint algorithm that will take into account what needs to be moved, and work recursively somehow. This is a very rough idea, but it is an idea. Graphs? Model the game states as a graph and apply some sort of variation on a coloring algorithm, to resolve dependencies? Again, this is a very rough idea. A friend suggested genetic algorithms. This is sort of possible but not easily. I can't think of a good way to make an evaluation function, and without that we've got nothing. So the question is - How to create a program that takes a grid and the vehicle layout, and outputs a series of steps needed to get the red car out? Sub-issues: Finding some solution. Finding an optimal solution (minimal number of moves) Evaluating how good a current state is Example: How can you move the cars in this setting, so that the red car can "exit" the maze through the exit on the right?

  Read the article

• What algorithms are suitable for this simple machine learning problem?

  - by user213060

  I have a what I think is a simple machine learning question. Here is the basic problem: I am repeatedly given a new object and a list of descriptions about the object. For example: new_object: 'bob' new_object_descriptions: ['tall','old','funny']. I then have to use some kind of machine learning to find previously handled objects that had similar descriptions, for example, past_similar_objects: ['frank','steve','joe']. Next, I have an algorithm that can directly measure whether these objects are indeed similar to bob, for example, correct_objects: ['steve','joe']. The classifier is then given this feedback training of successful matches. Then this loop repeats with a new object. a Here's the pseudo-code: Classifier=new_classifier() while True: new_object,new_object_descriptions = get_new_object_and_descriptions() past_similar_objects = Classifier.classify(new_object,new_object_descriptions) correct_objects = calc_successful_matches(new_object,past_similar_objects) Classifier.train_successful_matches(object,correct_objects) But, there are some stipulations that may limit what classifier can be used: There will be millions of objects put into this classifier so classification and training needs to scale well to millions of object types and still be fast. I believe this disqualifies something like a spam classifier that is optimal for just two types: spam or not spam. (Update: I could probably narrow this to thousands of objects instead of millions, if that is a problem.) Again, I prefer speed when millions of objects are being classified, over accuracy. What are decent, fast machine learning algorithms for this purpose?

  Read the article

• Find optimal/good-enough strategy and AI for the game 'Proximity'?

  - by smci

  'Proximity' is a strategy game of territorial domination similar to Othello, Go and Risk. Two players, uses a 10x12 hex grid. Game invented by Brian Cable in 2007. Seems to be a worthy game for discussing a) optimal algorithm then b) how to build an AI. Strategies are going to be probabilistic or heuristic-based, due to the randomness factor, and the insane branching factor (20^120). So it will be kind of hard to compare objectively. A compute time limit of 5s per turn seems reasonable. Game: Flash version here and many copies elsewhere on the web Rules: here Object: to have control of the most armies after all tiles have been placed. Each turn you received a randomly numbered tile (value between 1 and 20 armies) to place on any vacant board space. If this tile is adjacent to any ally tiles, it will strengthen each tile's defenses +1 (up to a max value of 20). If it is adjacent to any enemy tiles, it will take control over them if its number is higher than the number on the enemy tile. Thoughts on strategy: Here are some initial thoughts; setting the computer AI to Expert will probably teach a lot: minimizing your perimeter seems to be a good strategy, to prevent flips and minimize worst-case damage like in Go, leaving holes inside your formation is lethal, only more so with the hex grid because you can lose armies on up to 6 squares in one move low-numbered tiles are a liability, so place them away from your main territory, near the board edges and scattered. You can also use low-numbered tiles to plug holes in your formation, or make small gains along the perimeter which the opponent will not tend to bother attacking. a triangle formation of three pieces is strong since they mutually reinforce, and also reduce the perimeter Each tile can be flipped at most 6 times, i.e. when its neighbor tiles are occupied. Control of a formation can flow back and forth. Sometimes you lose part of a formation and plug any holes to render that part of the board 'dead' and lock in your territory/ prevent further losses. Low-numbered tiles are obvious-but-low-valued liabilities, but high-numbered tiles can be bigger liabilities if they get flipped (which is harder). One lucky play with a 20-army tile can cause a swing of 200 (from +100 to -100 armies). So tile placement will have both offensive and defensive considerations. Comment 1,2,4 seem to resemble a minimax strategy where we minimize the maximum expected possible loss (modified by some probabilistic consideration of the value ß the opponent can get from 1..20 i.e. a structure which can only be flipped by a ß=20 tile is 'nearly impregnable'.) I'm not clear what the implications of comments 3,5,6 are for optimal strategy. Interested in comments from Go, Chess or Othello players. (The sequel ProximityHD for XBox Live, allows 4-player -cooperative or -competitive local multiplayer increases the branching factor since you now have 5 tiles in your hand at any given time, of which you can only play one. Reinforcement of ally tiles is increased to +2 per ally.)

  Read the article

• Neural Network settings for fast training

  - by danpalmer

  I am creating a tool for predicting the time and cost of software projects based on past data. The tool uses a neural network to do this and so far, the results are promising, but I think I can do a lot more optimisation just by changing the properties of the network. There don't seem to be any rules or even many best-practices when it comes to these settings so if anyone with experience could help me I would greatly appreciate it. The input data is made up of a series of integers that could go up as high as the user wants to go, but most will be under 100,000 I would have thought. Some will be as low as 1. They are details like number of people on a project and the cost of a project, as well as details about database entities and use cases. There are 10 inputs in total and 2 outputs (the time and cost). I am using Resilient Propagation to train the network. Currently it has: 10 input nodes, 1 hidden layer with 5 nodes and 2 output nodes. I am training to get under a 5% error rate. The algorithm must run on a webserver so I have put in a measure to stop training when it looks like it isn't going anywhere. This is set to 10,000 training iterations. Currently, when I try to train it with some data that is a bit varied, but well within the limits of what we expect users to put into it, it takes a long time to train, hitting the 10,000 iteration limit over and over again. This is the first time I have used a neural network and I don't really know what to expect. If you could give me some hints on what sort of settings I should be using for the network and for the iteration limit I would greatly appreciate it. Thank you!

  Read the article

• Designing bayesian networks

  - by devoured elysium

  I have a basic question about Bayesian networks. Let's assume we have an engine, that with 1/3 probability can stop working. I'll call this variable ENGINE. If it stops working, then your car doesn't work. If the engine is working, then your car will work 99% of the time. I'll call this one CAR. Now, if your car is old(OLD), instead of not working 1/3 of the time, your engine will stop working 1/2 of the time. I'm being asked to first design the network and then assign all the conditional probabilities associated with the table. I'd say the diagram of this network would be something like OLD -> ENGINE -> CAR Now, for the conditional probabilities tables I did the following: OLD |ENGINE ------------ True | 0.50 False | 0.33 and ENGINE|CAR ------------ True | 0.99 False | 0.00 Now, I am having trouble about how to define the probabilities of OLD. In my point of view, old is not something that has a CAUSE relationship with ENGINE, I'd say it is more a characteristic of it. Maybe there is a different way to express this in the diagram? If the diagram is indeed correct, how would I go to make the tables? Thanks

  Read the article

• Higher-order unification

  - by rwallace

  I'm working on a higher-order theorem prover, of which unification seems to be the most difficult subproblem. If Huet's algorithm is still considered state-of-the-art, does anyone have any links to explanations of it that are written to be understood by a programmer rather than a mathematician? Or even any examples of where it works and the usual first-order algorithm doesn't?

  Read the article

• chess AI for GAE

  - by Richard

  I am looking for a Chess AI that can be run on Google App Engine. Most chess AI's seem to be written in C and so can not be run on the GAE. It needs to be strong enough to beat a casual player, but efficient enough that it can calculate a move within a single request (less than 10 secs). Ideally it would be written in Python for easier integration with existing code. I came across a few promising projects but they don't look mature: http://code.google.com/p/chess-free http://mariobalibrera.com/mics/ai.html

  Read the article

• how to work with strings and integers as bit strings in python?

  - by Manuel

  Hello! I'm developing a Genetic Algorithm in python were chromosomes are composed of strings and integers. To apply the genetic operations, I want to convert these groups of integers and strings into bit strings. For example, if one chromosome is: ["Hello", 4, "anotherString"] I'd like it to become something like: 0100100100101001010011110011 (this is not actual translation). So... How can I do this? Chromosomes will contain the same amount of strings and integers, but this numbers can vary from one algorithm run to another. To be clear, what I want to obtain is the bit representation of each element in the chromosome concatenated. If you think this would not be the best way to apply genetic operators (such as mutation and simple crossover) just tell me! I'm open to new ideas. Thanks a lot! Manuel

  Read the article

• How to find minimum cut-sets for several subgraphs of a graph of degrees 2 to 4

  - by Tore

  I have a problem, Im trying to make A* searches through a grid based game like pacman or sokoban, but i need to find "enclosures". What do i mean by enclosures? subgraphs with as few cut edges as possible given a maximum size and minimum size for number of vertices that act as soft constraints. Alternatively you could say i am looking to find bridges between subgraphs, but its generally the same problem. Given a game that looks like this, what i want to do is find enclosures so that i can properly find entrances to them and thus get a good heuristic for reaching vertices inside these enclosures. So what i want is to find these colored regions on any given map. The reason for me bothering to do this and not just staying content with the performance of a simple manhattan distance heuristic is that an enclosure heuristic can give more optimal results and i would not have to actually do the A* to get some proper distance calculations and also for later adding competitive blocking of opponents within these enclosures when playing sokoban type games. Also the enclosure heuristic can be used for a minimax approach to finding goal vertices more properly. Do you know of a good algorithm for solving this problem or have any suggestions in things i should explore?

  Read the article

• Neural Network with softmax activation

  - by Cambium

  This is more or less a research project for a course, and my understanding of NN is very/fairly limited, so please be patient :) ============== I am currently in the process of building a neural network that attempts to examine an input dataset and output the probability/likelihood of each classification (there are 5 different classifications). Naturally, the sum of all output nodes should add up to 1. Currently, I have two layers, and I set the hidden layer to contain 10 nodes. I came up with two different types of implementations 1) Logistic sigmoid for hidden layer activation, softmax for output activation 2) Softmax for both hidden layer and output activation I am using gradient descent to find local maximums in order to adjust the hidden nodes' weights and the output nodes' weights. I am certain in that I have this correct for sigmoid. I am less certain with softmax (or whether I can use gradient descent at all), after a bit of researching, I couldn't find the answer and decided to compute the derivative myself and obtained softmax'(x) = softmax(x) - softmax(x)^2 (this returns an column vector of size n). I have also looked into the MATLAB NN toolkit, the derivative of softmax provided by the toolkit returned a square matrix of size nxn, where the diagonal coincides with the softmax'(x) that I calculated by hand; and I am not sure how to interpret the output matrix. I ran each implementation with a learning rate of 0.001 and 1000 iterations of back propagation. However, my NN returns 0.2 (an even distribution) for all five output nodes, for any subset of the input dataset. My conclusions: o I am fairly certain that my gradient of descent is incorrectly done, but I have no idea how to fix this. o Perhaps I am not using enough hidden nodes o Perhaps I should increase the number of layers Any help would be greatly appreciated! The dataset I am working with can be found here (processed Cleveland): http://archive.ics.uci.edu/ml/datasets/Heart+Disease

  Read the article

• Update Rule in Temporal difference

  - by Betamoo

  The update rule TD(0) Q-Learning: Q(t-1) = (1-alpha) * Q(t-1) + (alpha) * (Reward(t-1) + gamma* Max( Q(t) ) ) Then take either the current best action (to optimize) or a random action (to explorer) Where MaxNextQ is the maximum Q that can be got in the next state... But in TD(1) I think update rule will be: Q(t-2) = (1-alpha) * Q(t-2) + (alpha) * (Reward(t-2) + gamma * Reward(t-1) + gamma * gamma * Max( Q(t) ) ) My question: The term gamma * Reward(t-1) means that I will always take my best action at t-1 .. which I think will prevent exploring.. Can someone give me a hint? Thanks

  Read the article

• mixture of expert

  - by Hamid

  Hi everybody, I'm looking for mixture of MLP(Multilayer Perceptron) expert matlab's code.could you please help me how i can find it? Thanks

  Read the article

• Using Hidden Markov Model for designing AI mp3 player

  - by Casper Slynge

  Hey guys. Im working on an assignment, where I want to design an AI for a mp3 player. The AI must be trained and designed with the use of a HMM method. The mp3 player shall have the functionality of adapting to its user, by analyzing incoming biological sensor data, and from this data the mp3 player will choose a genre for the next song. Given in the assignment is 14 samples of data: One sample consist of Heart Rate, Respiration, Skin Conductivity, Activity and finally the output genre. Below is the 14 samples of data, just for you to get an impression of what im talking about. Sample HR RSP SC Activity Genre S1 Medium Low High Low Rock S2 High Low Medium High Rock S3 High High Medium Low Classic S4 High Medium Low Medium Classic S5 Medium Medium Low Low Classic S6 Medium Low High High Rock S7 Medium High Medium Low Classic S8 High Medium High Low Rock S9 High High Low Low Classic S10 Medium Medium Medium Low Classic S11 Medium Medium High High Rock S12 Low Medium Medium High Classic S13 Medium High Low Low Classic S14 High Low Medium High Rock My time of work regarding HMM is quite low, so my question to you is if I got the right angle on the assignment. I have three different states for each sensor: Low, Medium, High. Two observations/output symbols: Rock, Classic In my own opinion I see my start probabilities as the weightened factors for either a Low, Medium or High state in the Heart Rate. So the ideal solution for the AI is that it will learn these 14 sets of samples. And when a users sensor input is received, the AI will compare the combination of states for all four sensors, with the already memorized samples. If there exist a matching combination, the AI will choose the genre, and if not it will choose a genre according to the weightened transition probabilities, while simultaniously updating the transition probabilities with the new data. Is this a right approach to take, or am I missing something ? Is there another way to determine the output probability (read about Maximum likelihood estimation by EM, but dont understand the concept)? Best regards, Casper

  Read the article

• A* (A-star) implementation in AS3

  - by Bryan Hare

  Hey, I am putting together a project for a class that requires me to put AI in a top down Tactical Strategy game in Flash AS3. I decided that I would use a node based path finding approach because the game is based on a circular movement scheme. When a player moves a unit he essentially draws a series of line segments that connect that a player unit will follow along. I am trying to put together a similar operation for the AI units in our game by creating a list of nodes to traverse to a target node. Hence my use of Astar (the resulting path can be used to create this line). Here is my Algorithm function findShortestPath (startN:node, goalN:node) { var openSet:Array = new Array(); var closedSet:Array = new Array(); var pathFound:Boolean = false; startN.g_score = 0; startN.h_score = distFunction(startN,goalN); startN.f_score = startN.h_score; startN.fromNode = null; openSet.push (startN); var i:int = 0 for(i= 0; i< nodeArray.length; i++) { for(var j:int =0; j<nodeArray[0].length; j++) { if(!nodeArray[i][j].isPathable) { closedSet.push(nodeArray[i][j]); } } } while (openSet.length != 0) { var cNode:node = openSet.shift(); if (cNode == goalN) { resolvePath (cNode); return true; } closedSet.push (cNode); for (i= 0; i < cNode.dirArray.length; i++) { var neighborNode:node = cNode.nodeArray[cNode.dirArray[i]]; if (!(closedSet.indexOf(neighborNode) == -1)) { continue; } neighborNode.fromNode = cNode; var tenativeg_score:Number = cNode.gscore + distFunction(neighborNode.fromNode,neighborNode); if (openSet.indexOf(neighborNode) == -1) { neighborNode.g_score = neighborNode.fromNode.g_score + distFunction(neighborNode,cNode); if (cNode.dirArray[i] >= 4) { neighborNode.g_score -= 4; } neighborNode.h_score=distFunction(neighborNode,goalN); neighborNode.f_score=neighborNode.g_score+neighborNode.h_score; insertIntoPQ (neighborNode, openSet); //trace(" F Score of neighbor: " + neighborNode.f_score + " H score of Neighbor: " + neighborNode.h_score + " G_score or neighbor: " +neighborNode.g_score); } else if (tenativeg_score <= neighborNode.g_score) { neighborNode.fromNode=cNode; neighborNode.g_score=cNode.g_score+distFunction(neighborNode,cNode); if (cNode.dirArray[i]>=4) { neighborNode.g_score-=4; } neighborNode.f_score=neighborNode.g_score+neighborNode.h_score; openSet.splice (openSet.indexOf(neighborNode),1); //trace(" F Score of neighbor: " + neighborNode.f_score + " H score of Neighbor: " + neighborNode.h_score + " G_score or neighbor: " +neighborNode.g_score); insertIntoPQ (neighborNode, openSet); } } } trace ("fail"); return false; } Right now this function creates paths that are often not optimal or wholly inaccurate given the target and this generally happens when I have nodes that are not path able, and I am not quite sure what I am doing wrong right now. If someone could help me correct this I would appreciate it greatly. Some Notes My OpenSet is essentially a Priority Queue, so thats how I sort my nodes by cost. Here is that function function insertIntoPQ (iNode:node, pq:Array) { var inserted:Boolean=true; var iterater:int=0; while (inserted) { if (iterater==pq.length) { pq.push (iNode); inserted=false; } else if (pq[iterater].f_score >= iNode.f_score) { pq.splice (iterater,0,iNode); inserted=false; } ++iterater; } } Thanks!

  Read the article

• Pong: How does the paddle know where the ball will hit?

  - by Roflcoptr

  After implementing Pacman and Snake I'm implementing the next very very classic game: Pong. The implementation is really simple, but I just have one little problem remaining. When one of the paddle (I'm not sure if it is called paddle) is controlled by the computer, I have trouble to position it at the correct position. The ball has a current position, a speed (which for now is constant) and a direction angle. So I could calculate the position where it will hit the side of the computer controlled paddle. And so Icould position the paddle right there. But however in the real game, there is a probability that the computer's paddle will miss the ball. How can I implement this probability? If I only use a probability of lets say 0.5 that the computer's paddle will hit the ball, the problem is solved, but I think it isn't that simple. From the original game I think the probability depends on the distance between the current paddle position and the position the ball will hit the border. Does anybody have any hints how exactly this is calculated?

  Read the article

• Quantum PSO and Charged PSO (PSO = Particle Swarm Optimizer)

  - by The Elite Gentleman

  Hi Guys I need to implement PSO's (namely charged and quantum PSO's). My questions are these: What Velocity Update strategy do each PSO's use (Synchronous or Asynchronous particle update) What social networking topology does each of the PSO's use (Von Neumann, Ring, Star, Wheel, Pyramid, Four Clusters) For now, these are my issues. All your help will be appreciated. Thanks.

  Read the article

• Neural Network: Handling unavailable inputs (missing or incomplete data)

  - by Mike

  Hopefully the last NN question you'll get from me this weekend, but here goes :) Is there a way to handle an input that you "don't always know"... so it doesn't affect the weightings somehow? Soo... if I ask someone if they are male or female and they would not like to answer, is there a way to disregard this input? Perhaps by placing it squarely in the centre? (assuming 1,0 inputs at 0.5?) Thanks

  Read the article

• Annoying Captcha >> How to programm a form that can SMELL difference between human and robot?

  - by Sam

  Hi folks. On the comment of my old form needing a CAPTHA, I felt I share my problem, perhaps you recognize it and find its time we had better solutions: FACTUAL PROBLEM I know most of my clients (typical age= 40~60) hate CAPTCHA things. Now, I myself always feel like a robot, when I have to sueeze my eyes and fill in the strange letters from the Capcha... Sometimes I fail! Go back etc. Turnoff. I mean comon its 2011, shouldnt the forms have better A.I. by now? MY NEW IDEA (please dont laugh) Ive thought about it and this is my idea's to tell difference between human and robot: My idea is to give credibility points. 100 points = human 0% = robot. require real human mouse movements require mousemovements that dont follow any mathematical pattern require non-instantaneous reading delays, between load and first input in form when typing in form, delays are measured between letters and words approve as human when typical human behaviour measured (deleting, rephrasing etc) dont allow instant pasting or all fields give points for real keyboard pressures retract points for credibility when hyperlinks in form Test wether fake email field (invisible by human) is populated (suggested by Tomalak) when more than 75% human cretibility, allow to be sent without captcha when less than 25% human crecibility, force captcha puzzle to be sure Could we write a A.I. PHP that replaces the human-annoying capthas, meanwhile stopping most spamservers filing in the data? Not only for the fun of it, but also actually to provide a 99% better alternative than CAPTHCA's. Imagine the userfriendlyness of your forms! Your site distinguishing itself from others, showing your audience your sites KNOWS the difference between a robot and a human. Imagine the advangage. I am trying to capture the essense of that distinguishing edge. PROGRAMMING QUESTION: 1) Are such things possible to programm? 2) If so how would you start such programm? 3) Are there already very good working solutions available elsewhere? 4) If it isn't so hard, your are welcome to share your answer/solutions below. 5) upon completion of hints and new ideas, could this page be the start of a new AI captcha, OR should I forget about it and just go with the flow, forget about the whole AI dream, and use captcha like everyone else.

  Read the article

• How can I make this Java code run faster?

  - by Martin Wiboe

  Hello all, I am trying to make a Java port of a simple feed-forward neural network. This obviously involves lots of numeric calculations, so I am trying to optimize my central loop as much as possible. The results should be correct within the limits of the float data type. My current code looks as follows (error handling & initialization removed): /** * Simple implementation of a feedforward neural network. The network supports * including a bias neuron with a constant output of 1.0 and weighted synapses * to hidden and output layers. * * @author Martin Wiboe */ public class FeedForwardNetwork { private final int outputNeurons; // No of neurons in output layer private final int inputNeurons; // No of neurons in input layer private int largestLayerNeurons; // No of neurons in largest layer private final int numberLayers; // No of layers private final int[] neuronCounts; // Neuron count in each layer, 0 is input // layer. private final float[][][] fWeights; // Weights between neurons. // fWeight[fromLayer][fromNeuron][toNeuron] // is the weight from fromNeuron in // fromLayer to toNeuron in layer // fromLayer+1. private float[][] neuronOutput; // Temporary storage of output from previous layer public float[] compute(float[] input) { // Copy input values to input layer output for (int i = 0; i < inputNeurons; i++) { neuronOutput[0][i] = input[i]; } // Loop through layers for (int layer = 1; layer < numberLayers; layer++) { // Loop over neurons in the layer and determine weighted input sum for (int neuron = 0; neuron < neuronCounts[layer]; neuron++) { // Bias neuron is the last neuron in the previous layer int biasNeuron = neuronCounts[layer - 1]; // Get weighted input from bias neuron - output is always 1.0 float activation = 1.0F * fWeights[layer - 1][biasNeuron][neuron]; // Get weighted inputs from rest of neurons in previous layer for (int inputNeuron = 0; inputNeuron < biasNeuron; inputNeuron++) { activation += neuronOutput[layer-1][inputNeuron] * fWeights[layer - 1][inputNeuron][neuron]; } // Store neuron output for next round of computation neuronOutput[layer][neuron] = sigmoid(activation); } } // Return output from network = output from last layer float[] result = new float[outputNeurons]; for (int i = 0; i < outputNeurons; i++) result[i] = neuronOutput[numberLayers - 1][i]; return result; } private final static float sigmoid(final float input) { return (float) (1.0F / (1.0F + Math.exp(-1.0F * input))); } } I am running the JVM with the -server option, and as of now my code is between 25% and 50% slower than similar C code. What can I do to improve this situation? Thank you, Martin Wiboe

  Read the article

• how useful is Turing completeness? are neural nets turing complete?

  - by Albert

  While reading some papers about the Turing completeness of recurrent neural nets (for example: Turing computability with neural nets, Hava T. Siegelmann and Eduardo D. Sontag, 1991), I got the feeling that the proof which was given there was not really that practical. For example the referenced paper needs a neural network which neuron activity must be of infinity exactness (to reliable represent any rational number). Other proofs need a neural network of infinite size. Clearly, that is not really that practical. But I started to wonder now if it does make sense at all to ask for Turing completeness. By the strict definition, no computer system nowadays is Turing complete because none of them will be able to simulate the infinite tape. Interestingly, programming language specification leaves it most often open if they are turing complete or not. It all boils down to the question if they will always be able to allocate more memory and if the function call stack size is infinite. Most specification don't really specify this. Of course all available implementations are limited here, so all practical implementations of programming languages are not Turing complete. So, what you can say is that all computer systems are just equally powerful as finite state machines and not more. And that brings me to the question: How useful is the term Turing complete at all? And back to neural nets: For any practical implementation of a neural net (including our own brain), they will not be able to represent an infinite number of states, i.e. by the strict definition of Turing completeness, they are not Turing complete. So does the question if neural nets are Turing complete make sense at all? The question if they are as powerful as finite state machines was answered already much earlier (1954 by Minsky, the answer of course: yes) and also seems easier to answer. I.e., at least in theory, that was already the proof that they are as powerful as any computer.

  Read the article

• Artifical neural networks height-weight problem

  - by hammid1981

  i plan to use neurodotnet for my phd thesis, but before that i just want to build some small solutions to get used to the dll structure. the first problem that i want to model using backward propagation is height-weight ratio. I have some height and weight data, i want to train my NN so that if i put in some weight then i should get correct height as a output. i have 1 input 1 hidden and 1 output layer. Now here is first of many things i cant get around :) 1. my height data is in form of 1.422, 1.5422 ... etc and the corresponding weight data is 90 95, but the NN takes the input as 0/1 or -1/1 and given the output in the same range. how to address this problem

  Read the article

• Prolog: Not executing code as expected.

  - by Louis

  Basically I am attempting to have an AI agent navigate a world based on given percepts. My issue is handling how the agent moves. Basically, I have created find_action/4 such that we pass in the percepts, action, current cell, and the direction the agent is facing. As it stands the entire code looks like: http://wesnoth.pastebin.com/kdNvzZ6Y My issue is mainly with lines 102 to 106. Basically, in it's current form the code does not work and the find_action is skipped even when the agent is in fact facing right (I have verified this). This broken code is as follows: % If we are headed right, take a left turn find_action([_, _, _, _, _], Action, _, right) :- retractall(facing(_)), assert(facing(up)), Action = turnleft . However, after some experimentation I have concluded that the following works: % If we are headed right, take a left turn find_action([_, _, _, _, _], Action, _, _) :- facing(right), retractall(facing(_)), assert(facing(up)), Action = turnleft . I am not entire sure why this is. I've attempted to create several identical find_action's as well, each checking a different direction using the facing(_) format, however swipl does not like this and throws an error. Any help would be greatly appreciated.

  Read the article

• TicTacToe AI Making Incorrect Decisions

  - by Chris Douglass

  A little background: as a way to learn multinode trees in C++, I decided to generate all possible TicTacToe boards and store them in a tree such that the branch beginning at a node are all boards that can follow from that node, and the children of a node are boards that follow in one move. After that, I thought it would be fun to write an AI to play TicTacToe using that tree as a decision tree. TTT is a solvable problem where a perfect player will never lose, so it seemed an easy AI to code for my first time trying an AI. Now when I first implemented the AI, I went back and added two fields to each node upon generation: the # of times X will win & the # of times O will win in all children below that node. I figured the best solution was to simply have my AI on each move choose and go down the subtree where it wins the most times. Then I discovered that while it plays perfect most of the time, I found ways where I could beat it. It wasn't a problem with my code, simply a problem with the way I had the AI choose it's path. Then I decided to have it choose the tree with either the maximum wins for the computer or the maximum losses for the human, whichever was more. This made it perform BETTER, but still not perfect. I could still beat it. So I have two ideas and I'm hoping for input on which is better: 1) Instead of maximizing the wins or losses, instead I could assign values of 1 for a win, 0 for a draw, and -1 for a loss. Then choosing the tree with the highest value will be the best move because that next node can't be a move that results in a loss. It's an easy change in the board generation, but it retains the same search space and memory usage. Or... 2) During board generation, if there is a board such that either X or O will win in their next move, only the child that prevents that win will be generated. No other child nodes will be considered, and then generation will proceed as normal after that. It shrinks the size of the tree, but then I have to implement an algorithm to determine if there is a one move win and I think that can only be done in linear time (making board generation a lot slower I think?) Which is better, or is there an even better solution?

  Read the article

• How do I create a good evaluation function for a new board game?

  - by A. Rex

  I write programs to play board game variants sometimes. The basic strategy is standard alpha-beta pruning or similar searches, sometimes augmented by the usual approaches to endgames or openings. I've mostly played around with chess variants, so when it comes time to pick my evaluation function, I use a basic chess evaluation function. However, now I am writing a program to play a completely new board game. How do I choose a good or even decent evaluation function? The main challenges are that the same pieces are always on the board, so a usual material function won't change based on position, and the game has been played less than a thousand times or so, so humans don't necessarily play it enough well yet to give insight. (PS. I considered a MoGo approach, but random games aren't likely to terminate.) Any ideas? Game details: The game is played on a 10-by-10 board with a fixed six pieces per side. The pieces have certain movement rules, and interact in certain ways, but no piece is ever captured. The goal of the game is to have enough of your pieces in certain special squares on the board. The goal of the computer program is to provide a player which is competitive with or better than current human players.

  Read the article

< Previous Page | 11 12 13 14 15 16 17 18 19 20 21 22  | Next Page >