pytorch
小然_ran
南京理工与信工所博士,喜欢机器学习,深度学习
展开
-
一种被AOD-Net启发的超高清图像去雾模型
GitHub - zzr-idam/AOD-Net_UHDdehazing原创 2022-02-26 14:26:50 · 717 阅读 · 0 评论 -
self-attention(pytorch 实现)
来源:MEF-GAN: Multi-Exposure Image Fusion via Generative Adversarial Networksclass Attention(nn.Module): def __init__(self, bn=True): super(Attention, self).__init__() self.conv1 = nn.Conv2d(6, 16, kernel_size=3, stride=2) ..原创 2020-12-13 17:41:03 · 6267 阅读 · 5 评论 -
深度网络与tucker分解结合实现数据分类(pytorch)
importtorchfromtensorly.decompositionimportnon_negative_tuckerimporttensorlyastlfromtensorlyimporttucker_to_tensorfromsklearn.clusterimportk_meansfromtensorly.tucker_tensorimporttucker_to_tensorfromtensorly.randomimportcheck...原创 2020-12-05 21:15:10 · 710 阅读 · 0 评论 -
pytorch实现简单的网络全局剪枝
importtorchfromtorchimportnnimporttorch.nn.utils.pruneaspruneimporttorch.nn.functionalasFdevice=torch.device("cuda"iftorch.cuda.is_available()else"cpu")classLeNet(nn.Module):def__init__(self):super(LeNet,s...原创 2020-08-15 11:17:51 · 993 阅读 · 5 评论 -
在MNIST数据集上实现特征可视化技术(CAM&pytorch)
importtorchimporttorch.nnasnnimporttorch.nn.functionalasFimporttorch.optimasoptimimporttorch.utils.dataasDatafromtorchvisionimportdatasets,transformsfromtorch.autogradimportVariableBATCH_SIZE=64train_dataset=...原创 2020-08-05 16:04:57 · 756 阅读 · 0 评论 -
GAN网络生成鸢尾花样本(pytorch)
importtorchimporttorch.nnasnnimporttorch.optimasoptimfromsklearnimportdatasets#潜在的空间其实GAN训练出来的判别器对对抗样例的防御是有意义的。但是#为了防御对抗样例而训练代价太大了。#z潜在mapnz=10#生成器mapngf=20#判别器mapndf=20lr=0.001#X=to...原创 2020-08-04 12:31:57 · 1453 阅读 · 3 评论 -
利用鸢尾花数据集直观感受神经网络表征能力(pytorch)
importtorchimporttorchimporttorchvisionfromtorch.autogradimportVariablefromtorchvisionimportdatasets,transformsfromtorch.utils.dataimportDataLoaderimportcv2fromtorchimportnnfromnumpyimport*importtorch.nn.functional...原创 2020-07-25 17:41:35 · 576 阅读 · 0 评论 -
利用高斯核卷积对MINIST数据集进行去噪
importtorchimporttorchvisionfromtorch.autogradimportVariablefromtorchvisionimportdatasets,transformsfromtorch.utils.dataimportDataLoaderimportcv2fromtorchimportnnimportnumpyasnpimporttorch.nn.functionalasFimportadve...原创 2020-07-18 16:24:30 · 1090 阅读 · 0 评论 -
特征矩阵稀疏化(pytorch)
importtorcha=torch.tensor([[1,1,1,1.0],[2,2,2,2],[1,1,1,1],[1,1,1,1]])b,indices=torch.nn.MaxPool2d(2,2,return_indices=True)(a.unsqueeze(0).unsqueeze(0))print(b...原创 2020-07-13 21:25:22 · 1414 阅读 · 0 评论 -
针对神经网络的快速符号攻击
'''但是,设计和训练模型的一个经常被忽略的方面是安全性和鲁棒性,尤其是在面对想要欺骗模型的对手的情况下这篇文章的思路就是给图像加噪音,但是大家会发现一件事,加噪音我们希望给神经网络认为重要的像素点给予更多的攻击,作者的思路是认为,根据梯度来反馈的到大小来反推像素点的重要性,一般来说能使神经网络获得很大梯度的像素点肯定是重要的。'''importtorchimporttorch.nnasnnim...原创 2020-06-19 21:49:54 · 577 阅读 · 1 评论 -
利用自编码器对线性模型参数加密
fromsklearn.datasetsimportload_bostondataset=load_boston()x_data=dataset.datay_data=dataset.targetfromsklearnimportlinear_modelmodel=linear_model.LinearRegression()model.fit(x_data,y_data)args=model.coef_im...原创 2020-06-18 09:24:08 · 332 阅读 · 0 评论 -
行为匹配(采用变分自编码器对数据降维)
importnumpyasnpimporttorchimporttorch.nnasnnimporttorch.nn.functionalasF#Cosinesimilaritydefcos_sim(a,b):numerator=np.multiply(a,b).sum()denomitor=np.lina...原创 2020-02-15 10:14:55 · 1065 阅读 · 0 评论 -
深度神经网络炼丹与优化方法
链接:https://pan.baidu.com/s/1o-GwTf9mKctEbkv-jA8TqQ提取码:cvb8原创 2020-01-23 17:08:38 · 222 阅读 · 0 评论 -
借助于pytorch的GNN网络解决鸢尾花分类问题
from sklearn.datasets import load_irisimport torchfrom torch_geometric.data import Dataimport networkx as nximport torch.nn.functional as Ffrom torch_geometric.nn import GCNConvdata = load_iris...原创 2020-01-03 15:36:43 · 1643 阅读 · 1 评论 -
图像模糊聚类
1. 用池化操作对图像降维打击2. 利用混合高斯模型分块降维import torchvisionfrom PIL import Imageimport torch.nn.functional as Floader = torchvision.transforms.Compose( [torchvision.transforms.Resize(128), ...原创 2019-12-30 14:52:03 · 441 阅读 · 0 评论 -
针对复合函数的LSTM时序预测
import torchfrom torch.autograd import Variableimport torch.nn.functional as Fimport matplotlib.pyplot as pltimport torch.nn as nnimport mathimport numpy as npimport matplotlib.pyplot as plti...原创 2019-04-09 00:12:18 · 229 阅读 · 0 评论 -
输出各层神经网络权重与偏置
from torch import nnimport torch.nn.functional as fclass SimpleNet(nn.Module): def __init__(self, in_dim, n_hidden_1, n_hidden_2, out_dim): super(SimpleNet, self).__init__() self...原创 2019-05-29 17:15:09 · 3325 阅读 · 0 评论 -
基于pytorch岭回归
import torch.nnimport torch.optimx = torch.tensor([[1., 1., 1.], [2., 3., 1.], [3., 5., 1.], [4., 2., 1.], [5., 4., 1.]])y = torch.tensor([-10., 12., 14., 16., 18.])w = torch.zeros(3, requires_gr...原创 2019-06-26 09:30:45 · 1358 阅读 · 3 评论 -
贝叶斯网络对手写识别体的预测
import pyvarinffrom torch.autograd import Variableimport torch.nn as nnimport torch.nn.functional as Fimport torch.optim as optimfrom sklearn import datasetsimport torchfrom sklearn.model_sele...原创 2019-07-11 10:06:03 · 482 阅读 · 0 评论 -
pytorch模型的数据初始化代码
import torch.nn as nnimport torch.nn.functional as Fdef initialize_weights(self): for m in self.modules(): if isinstance(m, nn.Linear): nn.init.normal_(m.weight.data, 0, 0....原创 2019-07-15 17:18:42 · 703 阅读 · 0 评论 -
互相关操作与卷积操作
import numpy as npimport torchdata = np.correlate([1, 2], [1, 2], "full")print(data)x = torch.Tensor([[1, 2]])kernel = torch.Tensor([[1, 2]]).unsqueeze(0).unsqueeze(0).float()weight = torch.nn.P...原创 2019-05-17 20:18:32 · 1178 阅读 · 0 评论 -
基于pytorch线性判决器(不使用nn.moudle)
import torchimport matplotlib.pyplot as pltfrom sklearn.model_selection import train_test_splitfrom sklearn import datasetsfrom torch.utils.data import TensorDataset, DataLoaderdigits = datasets.load...原创 2019-05-17 17:16:29 · 212 阅读 · 0 评论 -
基于pytorch, softmax,logsoftmax 表达
import torchimport numpy as npinput = torch.autograd.Variable(torch.rand(1, 3))print(input)print('softmax={}'.format(torch.nn.functional.softmax(input, dim=1)))print('logsoftmax={}'.format(...原创 2019-05-27 09:47:53 · 1612 阅读 · 0 评论 -
基于LSTM的时间序列填充缺失值方法(pytorch)
import torchclass RNN(torch.nn.Module): def __init__(self): super().__init__() self.rnn = torch.nn.LSTM( input_size=1, hidden_size=64, num_...原创 2019-05-04 10:00:11 · 9219 阅读 · 3 评论 -
基于pytorch全连接神经网络手写体数据识别,准确率达到百分之97
import torchfrom torch import nnimport torch.optim as optimizerfrom torch.autograd import Variableimport matplotlib.pyplot as pltfrom sklearn.model_selection import train_test_splitfrom sklear...原创 2019-05-03 13:14:30 · 958 阅读 · 0 评论 -
基于pytorch的逻辑回归代码
from torch import nnfrom torch import optim as optimizerimport torchfrom torch.autograd import Variableclass LogisticRegression(nn.Module): def __init__(self): super(LogisticRegressi...原创 2019-05-02 21:05:47 · 319 阅读 · 0 评论 -
基于pytorch的线性回归算法并比较了Adam与SGD的收敛速度
from torch import nnfrom torch import optim as optimizerimport torchfrom torch.autograd import Variableimport matplotlib.pyplot as pltimport numpy as npclass Model(nn.Module): def __init__...原创 2019-05-02 12:37:53 · 1067 阅读 · 0 评论 -
pytorch简单实现GAN的分类
github网址:https://github.com/ranran4082391/GAN原创 2019-04-06 09:24:20 · 2023 阅读 · 5 评论 -
pytorch 搭建的基于LSTM自编码器对数据降维并采用KNN算法对鸢尾花分类
LSTM搭建自编码器提取特征,KNN分类import torchimport torch.nn as nnfrom sklearn.datasets import load_irisfrom sklearn.model_selection import train_test_split# 超参数EPOCH = 200LR = 0.005data = load_iris()...原创 2019-03-20 16:43:56 · 5253 阅读 · 10 评论 -
自编码器对数据降维,最后采用LSTM对数据进行拟合预测
采用自编码器对高维特征表达成低维特征,最后采用LSTM对数据进行拟合,预测准确率很高。import torchimport torch.nn as nnfrom sklearn.datasets import load_irisimport matplotlib.pyplot as pltimport numpy as npfrom sklearn.model_selection ...原创 2019-03-16 10:25:10 · 7657 阅读 · 13 评论 -
统计学习方法求PI(蒙特卡罗算法和拉斯维加斯算法)
1.蒙特卡罗算法计算圆周率,蒙特卡洛算法在相对确定的时间里获得可能不完全正确的结果。但是从概率上看,运行的时间越长,获得结果越可能越可能正确,或和正确值越接近。2.拉斯维加斯算法,拉斯维加斯算法总是能获得正确的结果,但是运行时间不定。https://github.com/ranran4082391/xiaoran_5...原创 2019-03-04 18:17:39 · 364 阅读 · 0 评论 -
基于pytorch的神经网络训练多标记数据
from sklearn.datasets import make_classificationimport torchfrom torch.autograd import Variablefrom torch import nnfrom sklearn.utils import shuffleimport numpy as npX, y1 = make_classificati...原创 2019-05-08 11:40:57 · 808 阅读 · 0 评论 -
基于pytorch的CNN识别sklearn自带手写体数据,准确率贼高
import torchfrom torch.autograd import Variableimport matplotlib.pyplot as pltfrom sklearn import datasetsfrom torch import nnfrom sklearn.model_selection import train_test_splitdigits = datas...原创 2019-05-06 12:03:20 · 1245 阅读 · 0 评论 -
基于pytorch比较sigmoid和relu函数梯度下降速度
import torchimport torch.nn as nnfrom torch.autograd import Variableimport torch.optimimport matplotlib.pyplot as pltX = torch.Tensor([[1, 2], [2, 2], [2, 1], [1, 1], [4, 4], [4, 5], [5, 4], [...原创 2019-05-21 21:19:46 · 484 阅读 · 0 评论 -
感知机根据不同分布数据样本得到不同的损失值(基于pytorch),简单理解算法收敛性问题
# 测试函数最常用来解决的一个问题 :算法最后结果是不是收敛的'''比如我们想用一条线去 分割一个数据集 将他分成两个类别-----》这里我们用感知机算法来测试一下''''''感知机一般只能解决线性分类问题也就是说图像是这个样子的,我们目的是找到一条线讲他们分开'''import matplotlib.pyplot as pltimport numpy as npX = [[1...原创 2019-05-24 21:00:54 · 480 阅读 · 0 评论 -
基于pytorch实现牛顿迭代法求函数极小值
import torchfrom torch.autograd import Variabledef f(x): y = x ** 2 return yx = Variable(torch.Tensor([5]), requires_grad=True)for i in range(100): grad_x = torch.autograd.grad(f(x)...原创 2019-05-20 17:23:14 · 1771 阅读 · 0 评论 -
基于pytorch求自构造函数的高阶导数
import torchfrom torch.autograd import Variabledef f(x): y = x ** 2 return yx = Variable(torch.Tensor([5]), requires_grad=True)grad_x = torch.autograd.grad(f(x), x, create_graph=True)p...原创 2019-05-20 17:00:14 · 1158 阅读 · 0 评论 -
基于pytorch手动code损失函数,并比较了7种梯度下降算法
import torchimport matplotlib.pyplot as pltimport torch.optim as optimimport randomdef grad_down(optimizer, x, w, y): optimizer.zero_grad() output = torch.mv(x, w) loss = MSELoss(out...原创 2019-05-20 11:55:20 · 442 阅读 · 0 评论 -
基于pytorch多标记实例,实现自动对两个未知数的求和与求积
import torchfrom torch.autograd import Variablefrom sklearn import datasetsfrom torch import nnX = [[i, i] for i in range(0, 10)]Y = [[i[0]*2, i[0]*i[0]] for i in X]X_train = Variable(torc...原创 2019-05-07 16:48:03 · 319 阅读 · 0 评论 -
残差神经网络训练sklearn手写体数据集(pytorch)
import torchfrom torch.autograd import Variablefrom sklearn import datasetsfrom torch import nndigits = datasets.load_digits()X = digits.imagesy = digits.targetX_train = X[:1700, :, :]Y_t...原创 2019-05-07 13:01:49 · 861 阅读 · 0 评论