written-Digits-Classification-using-MNIST-dataset
所属分类：模式识别(视觉/语音等)
开发工具：Python
文件大小：5KB
下载次数：0
上传日期：2018-04-20 02:24:57
上 传 者：
sh-1993
说明： 开发了一个5层序列卷积神经网络，使用Keras和Tensorflow后端进行数字识别...
(Developed a 5-layer Sequential Convolutional Neural Network using Keras with Tensorflow backend for digit recognition trained on MNIST dataset. Adjusted parameters such as kernel size, activation function and optimizer properties to compute the best fit. Obtained an accuracy of 97.12%. Performed Data Augmentation such as image scaling, image)
文件列表:
MNIST.py (7506, 2018-04-20)
MNIST_othermodels.py (3284, 2018-04-20)
# Handwritten-Digits-Classification-using-MNIST-dataset
Developed a 5-layer Sequential Convolutional Neural Network using Keras with Tensorflow backend for digit recognition trained on MNIST dataset. Adjusted parameters such as kernel size, activation function and optimizer properties to compute the best fit. Obtained an accuracy of 97.12%. Performed Data Augmentation such as image scaling, image flips and image rotation to avoid overfitting and increase the accuracy to ***.02%. Compared the accuracy to KNN, Logistic Regression, Random Forest which had accuracies of 96.37%, 91.22% and 96.19% respectively.
近期下载者：
相关文件：
收藏者：