GithubHelp home page GithubHelp logo

is-lab2's Introduction

IS-Lab2 (EN)

Intelligent systems. Task for the laboratory on Multi-Layer Perceptron training.

Aim

Learn to write training (parameter estimation) algorithm for the Multi-Layer Perceptron based approximator.

Tasks (maximum 8 points)

  1. Create a program to calculate the coefficients of the multilayer perceptron. The multilayer perceptron must perform the function of an approximator. Structure of the multilayer perceptron:
  • one input (input 20 input vectors (20 examples) X, with values in the range 0 to 1, eg x = 0.1: 1/22: 1;).
  • one output (for example, the output is expecting the desired response that can be calculated using the formula: y = (1 + 0.6 * sin (2 * pi * x / 0.7)) + 0.3 * sin (2 * pi * x)) / 2; - the neural network being created should "model / simulate the behavior of this formula" using a completely different mathematical expression than this);
  • One hidden layer with hyperbolic tangent or sigmoidal activation functions in neurons (number of neurons: 4-8);
  • linear activation function in the output neuron;
  • training algorithm - Backpropagation.

Additional task (2 additional points)

Solve the surface approximation task (two inputs and single output).

Suggested reading

  • Neural Networks and Learning Machines (3rd Edition), page <...>, Table 1.1.

IS-Lab2 (LT)

Intelektualiosios sistemos. Antrojo laboratorinio darbo užduotis.

Tikslas

Išmokti savarankiškai suprogramuoti paprasto netiesinio aproksimatoriaus mokymo (parametrų skaičiavimo) algoritmą.

Užduotys (maks. 8 balai)

  1. Sukurkite daugiasluoksnio perceptrono koeficientams apskaičiuoti skirtą programą. Daugiasluoksnis perceptronas turi atlikti aproksimatoriaus funkciją. Daugiasluoksnio perceptrono struktūra:
  • vienas įėjimas (įėjime paduodamas 20 skaičių vektorius X, su reikšmėmis intervale nuo 0 iki 1, pvz., x = 0.1:1/22:1; ).
  • vienas išėjimas (pvz., išėjime tikimasi tokio norimo atsako, kurį galima būtų apskaičiuoti pagal formulę: y = (1 + 0.6*sin(2*pi*x/0.7)) + 0.3*sin(2*pi*x))/2; - kuriamas neuronų tinklas turėtų "modeliuoti/imituoti šios formulės elgesį" naudodamas visiškai kitokią matematinę išraišką nei ši);
  • vienas paslėptasis sluoksnis su hiperbolinio tangento arba sigmoidinėmis aktyvavimo funkcijomis neuronuose (neuronų skaičius: 4-8);
  • tiesine aktyvavimo funkcija išėjimo neurone;
  • mokymo algoritmas - Backpropagation (atgalinio sklidimo).

Papildoma užduotis (papildomi 2 balai)

Išspręskite paviršiaus aproksimavimo uždavinį, kai tinklas turi du įėjimus ir vieną išėjimą.

Rekomenduojama literatūra

  • Neural Networks and Learning Machines (3rd Edition), <...> psl., <...> lentelė

is-lab2's People

Contributors

serackis avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.