Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/ksalekk/mlp-nn
https://github.com/ksalekk/mlp-nn
classification multilayer-perceptron neural-networks numpy python
Last synced: about 1 month ago
JSON representation
- Host: GitHub
- URL: https://github.com/ksalekk/mlp-nn
- Owner: ksalekk
- Created: 2024-05-22T13:01:33.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2024-05-30T23:34:30.000Z (9 months ago)
- Last Synced: 2024-11-20T12:36:29.933Z (3 months ago)
- Topics: classification, multilayer-perceptron, neural-networks, numpy, python
- Language: Python
- Homepage:
- Size: 158 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Multilayer Perceptron Neural Network
## Disclaimer
This app provides only information for educational purposes. This app is not medical or treatment advice, professional diagnosis, opinion, or services – and may not be treated as such by the user. As such, this app may not be relied upon for the purposes of medical diagnosis or as a recommendation for medical care or treatment. The information provided by this app is not a substitute for professional medical advice, diagnosis or treatment.
## General
Multilayer perceptron with backpropagation algorithm, implemented from stratch. Project developed for Neural Networks in Biomedicine course at the WUT.
## Architecture
Neural network is represented by *NeuralNetwork* object and it allows to maniupalate nn, e.g. set structure, learning params, start learning/testing process based on the specified dataset. *Layer* object represents the neural network layer (hidden or output) and allows to set layer params (inputs count, neurons/outputs count, activation function, bias). Every *Layer* object has weights matrix with *m x n* with weights for each input connection in the layer (*m-th* row contains *m-th* neuron inputs and *n-th* column contains *n-th* input in the layer).
```
BIAS IN_1 IN_2 ... IN_n
--- ---
NEURON_1 | w_1_0 w_1_1 w_1_2 ... w_1_n |
| |
NEURON_2 | w_2_0 w_2_1 w_2_2 ... w_2_n |
| |
... | ... ... ... ... ... |
| |
NEURON_m | w_3_0 w_m_1 w_m_2 ... w_m_n |
--- ---
```
## Dataset
The source of the data to train and test the network is [Elter,Matthias. (2007). Mammographic Mass. UCI Machine Learning Repository](https://doi.org/10.24432/C53K6Z). Missing cell values were filled with the median of the column in which the cell is missing. The dataset was divided in a proportion ~70:30 for traning and testing data.
## Results
For the structure given below and the learning rate of 0.4 the accuracy is ~0.75, the specificity ~0.71 and the sensitivity ~0.82.
Example error functions in the learning process are given below.
### Network scheme
### Example error functions
