搜索资源列表
RecognizeIt-v4.0-(softmax)
- 基于神经网络的手势识别程序 手写输入原理的演示-Gesture Recognition Based on Neural Network program demonstrates the principle of handwriting input
twodimapproximationbp
- 单输出函数Y=SIN(X)逼近问题的bp程序:假设网络结构为3--2--1,输入维数M,共N个样本,一般输入不算层,输出算层- 激活函数: hardlim---(0,1),hardlims---(-1,1),purelin,logsig---(0,1),tansig----(-1,1) softmax,poslin,radbas,satlin,satlins,tribas 训练算法: 1.traingd,traingdm,traingda(variable l
Exercise5-Softmax-Regression
- 斯坦福深度学习教程中关于softmax regression的练习代码,源代码中需要补全的地方,全部把代码补完整,把手写体识别的数据库放到路径下,可以直接运行-Stanford deep learning tutorial exercises on softmax regression code, source code need to fill all places, all the full complement of the code, the handwriting recognitio
UFLDL
- UFLDL作业,以下作业均已完整做好,build deep network,vectorization,softmax,working with large images,pca。-UFLDL,build deep network,vectorization,softmax,working with large images,pca
softmax-regression
- softmax 回归源代码(高维logistic回归 )进行邮件文本分类。-softmax regression to classify for email text.(high dimension of logistic regression)
libORF-master
- 针对各种机器学习,深度学习领域的一个matlab工具包-A machine learning library focused on deep learning.Following algorithms and models are provided along with some static utility classes: - Naive Bayes, Linear Regression, Logistic Regression, Softmax Regression, Linear S
stackedAE
- 堆栈自编码,通过两个稀疏自编码的堆叠和softmax分类模型,实现手写体的分类。-Stack self-encoding, since encoding by two sparse stack and softmax classification model to classify handwriting.
self-taught-learning
- 自主学习把稀疏自编码器和分类器实现结合。先通过稀疏自编码对无标签的5-9的手写体进行训练得到最优参数,然后通过前向传播,得到训练集和测试集的特征,通过0-4有标签训练集训练出softmax模型,然后输入测试集到分类模型实现分类。-Independent Learning the encoder and the sparse classifiers achieve the combination. First through sparse coding since no label was ha
Softmax_exercise
- Softmax用于多分类问题,本例是将MNIST手写数字数据库中的数据0-9十个数字进行分类,其中训练样本有6万个,测试样本有1万个数字是0~9-Softmax for multi classification problems, the present case is the handwritten data MNIST digital 0-9, classification, training samples which have 60,000, there are 10,000 test
stacked-autoencoder
- 基于两层的层叠自编码的深度学习模型,前两层用于特征提取,再加一个Softmax分类器用于分类-Two stacked the depth of learning coding model based on the first two levels for feature extraction, coupled with a classifier for classifying Softmax
Softmax
- 一个小demo,用softmax实现数据分类,数据有4类,均服从高斯分布-A small demo, a classifier with softmax softmax is implemented to classify 4 classes of datas which are generated by gaussian distribution
Classifier-based-Softmax
- 一个小demo,用softmax实现数据分类,数据有4类,均服从高斯分布-A little demo, a classifier with softmax softmax is implemented to classify 4 classes of datas which are generated by gaussian distribution
softmax
- 人工智能导论作业,用softmax方法实现的水杯图片分类,可扩展到其他分类任务(Homework of AI. Classify images of cups using softmax. Can be used in other tasks.)
mnist
- 利用keras实现手写数字识别,使用CNN模型 全连接层+两个卷积层,最后Softmax分类器,识别率超过96%(Using keras to realize handwritten numeral recognition baesd on CNN model. One whole connection layer + two convolution layers, and a Softmax classifier. The recognition accuracy is over 96%
TFFRCNN-master
- 识别给出的图像并用softmax给出可能的概率(Identify the given image and give possible probability with softmax.)
symbol_resnet
- RACNN注意力机制,细腻度图片分类。 RA-CNN由上到下用了3个尺度并且越来越精细,尺度间构成循环,即上层的输出作为当层的输入。RA-CNN主要包含两部分:每一个尺度上的卷积网络和相邻尺度间的注意力提取网络(APN, Attention Proposal Network)。在每一个尺度中,使用了堆叠的卷积层等,最后接上全连接层于softmax层,输出每一个类别的概率;这个是很好理解的,代码采用的网络结构是VGG的网络结构。(RACNN attention mechanism)
yolo V3
- 这个版本作者已经编译过了 可以在WIN系统上运行 YOLO3主要的改进有:调整了网络结构;利用多尺度特征进行对象检测;对象分类用Logistic取代了softmax。 在基本的图像特征提取方面,YOLO3采用了称之为Darknet-53的网络结构(含有53个卷积层),它借鉴了残差网络residual network的做法,在一些层之间设置了快捷链路(shortcut connections)。