Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Browse code

Added README in English

  • Loading branch information...
commit 2539ff16aabcdc67c88121ebb9e94af43a1e2018 1 parent 0a49c63
Mihai Maruseac authored

Showing 1 changed file with 36 additions and 0 deletions. Show diff stats Hide diff stats

  1. +36 0 README.rst
36 README.rst
Source Rendered
... ... @@ -0,0 +1,36 @@
  1 +Backpropagation
  2 +===============
  3 +
  4 +A. About
  5 +........
  6 +
  7 +This is a homework for the Machine Learning Course that I am taking right now
  8 +at my University.
  9 +
  10 +The homework uses backpropagation in order to forecast a value from a given set
  11 +of values, knowing that one value depends on the values of several of the
  12 +previous numbers in the series.
  13 +
  14 +The assignment is done in Python.
  15 +
  16 +B. Usage
  17 +........
  18 +
  19 +Run ``./bp.py`` to start the main part of the application. The GUI is pretty
  20 +simple to use.
  21 +
  22 +It will show a real time graph of the network while learning with edges
  23 +coloured according to their relevance: a red colour means inhibition while a
  24 +green color means a bonus.
  25 +
  26 +The application does a logging of all steps in the learning phase.
  27 +
  28 +The input is first normalized to the range of the activation function (-1 to 1
  29 +or 0 to 1). In fact, the normalization ensures that points outside the initial
  30 +range can still be somehow predicted (with a certain error if the data trend is
  31 +exponential but that is another problem).
  32 +
  33 +At the end of the learning phase, the user sees the predicted value along with
  34 +an estiamtion of the error and the application saves the network in different
  35 +formats to disk. Also, a plot of all data (predicted and given) is saved.
  36 +

0 comments on commit 2539ff1

Please sign in to comment.
Something went wrong with that request. Please try again.