文件列表:

MLP.gif (945275, 2020-05-14)
MLP.ipynb (13454, 2020-05-14)
MLP.jpg (28569, 2020-05-14)
MLP.py (5107, 2020-05-14)
View.py (4478, 2020-05-14)
banner.png (110225, 2020-05-14)
exmnist.png (137084, 2020-05-14)

# Neural-Network-Mnist-Dataset Repository dedicated to the classification of the minist database using neural networks programmed in python, only with the numpy library. ## Requirements - [Python3](https://www.python.org/downloads/release/python-382/) - *numpy* ```bash pip install numpy ``` - *progressbar* ```bash pip install progressbar ``` - *keras* ```bash pip install Keras ``` - *sklearn* ```bash pip install sklearn ``` ## Installation Clone this repository: ```bash https://github.com/SalesRyan/Neural-Network-MNIST-Dataset.git ``` ## Implementation ### [Dataset](https://www.tensorflow.org/datasets/catalog/mnist) The MNIST Data set is usually used to measure the efficiency of an algorithm in classifying images, so it was chosen to be the data set to be classified.

## Matrix Functions The cells presented in this section are responsible for operations between matrices and scales. ## MPL Implementation In this Section we have the implementation of the artificial neural network, with its functions, with emphasis on feedforward and backpropagation.

## Training The snippet of training code is very simple, only all training sets are shown to the network. ## Prediction The training code snippet happens the prediction of the test set, based on the training previously obtained. ## Validation The validation is composed by the metrics of accuracy recall that are usually used for validation of predictions. - [*Accuracy*](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.accuracy_score.html) - [*Recall*](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.recall_score.html) ## Conclusion The artificial neural network implemented in python only with the library numpy obtained a result of Accuracy: 0.93*** and Recal: 0.9400592307897068 for the MNITS dataset. ## Training view visualize the adjustment of weights during training. I programmed an interface with pygame, to facilitate visualization it was necessary to use a character resizing algorithm, I chose PCA (Principal Component Snalysis). You can check it out below. Run MLP.py.  ## About me Graduation in Information Systems at the Federal University of Piauí, Possibility of experience with developing solutions in the area of Digital Image Processing, Computer Vision and Artificial Intelligence. I currently participate in research projects in the area of Vision and Computational Intelligence. - [*LinkedIn*](https://www.linkedin.com/in/ryan-sales-2b10141a6/) - [*Lattes*](http://lattes.cnpq.br/694478106889***28) - [*Instagram*](https://www.instagram.com/sales.ryann/?hl=pt-br)

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：