GuokLiu的博客_CSDN博客-python,Pre-Knowledges,tensorflow领域博主

参考连接使用 Hyperopt 和 Plotly 可视化超参数优化Visualizing Hyperparameter Optimization with Hyperopt and Plotly — States Title源代码# plotly express不支持轮廓图，# 所以我们将使用'graph_objects'来代替。# `go.Contour`自动为我们的损失插入“z”值。fig = go.Figure( data=go.Contour( .

2022-05-31 09:06:25 303

原创 220526-PyTorch目标检测中的IoU计算

PyTorch 自带的IoU计算，只需提供两个检测框box1， box2的对角位置[X1, Y1, X2, Y2], [X3, Y3, X4, Y4]即可:记box1：左下坐标为[X1, Y1], 右上坐标为 [X2, Y2]记box1：左下坐标为[X3, Y3], 右上坐标为 [X4, Y4]J(A,B)=∣A∩B∣∣A∣∪∣B∣=∣A∣∩∣B∣∣A∣+∣B∣−∣A∩B∣J(A,B)=\frac{|A \cap B|}{|A| \cup |B|} = \frac{|A| \cap |B|}{.

2022-05-26 08:27:47 219

原创 220507-5步静态ip地址配置

Step 0: 命令行中ipconfig 查看ip地址Step 1 - 3: 各种点击Step 4: 双击IPv4Step 5: 配置静态ip地址Ip address: 填写Step 1中的前三个数字，最后一个数字设置为动态ip中不用的数字0-9？(这块不太清楚，就是选一个动态ip不用的数字)Subnet mask：点击自动生成，默认即可Default gateway: 同Step 1中前三个数字，最后一个设置成1Preferred DNS server：同Default gat.

2022-05-07 15:49:34 318

原创 220502-连续小波变换CWT中[scale, wavename]以及[contouf, pcolormesh]对信号分析的影响

Demo 1: CWT with different waveletsimport numpy as npimport matplotlib.pyplot as pltimport scipy.signal as signalimport pywtaa = []for i in range(200): aa.append(np.sin(0.3*np.pi*i))for i in range(200): aa.append(np.sin(0.13*np.pi*i))for i

2022-05-02 15:18:12 588

原创 220430-三种短时傅立叶变换STFT的绘制及数据提取

FigureCodeimport numpy as npimport matplotlib.pyplot as pltimport scipy.signal as signalsampling_rate = 8000.0fft_size = 1024step = fft_size/16time = 2t = np.arange(0, time, 1/sampling_rate)sweep = signal.chirp(t, f0=100, t1=time, f1=0.8*s.

2022-04-30 15:30:23 980

原创 220323-MacOS下的VsCode半透明主题设置

MacOS下的VsCode半透明主题设置参考： https://marketplace.visualstudio.com/items?itemName=eyhn.vscode-vibrancyhttps://marketplace.visualstudio.com/items?itemName=eyhn.vscode-vibrancyhttps://eyhn.in/vscode-vibrancy/

2022-03-23 16:37:17 1259

原创 220322-不同方法不同任务下的深度学习训练过程Train-Valid-Acc-Loss可视化

#%%import scipy.io as sioimport numpy as npimport matplotlib.pyplot as pltdef plot_curves(method, source, target, num_sam=10240, num_batch=256, num_epoch=200, top_epoch=20, ymin = -0.1, ymax=1.1, ax1=None): file = 'Exp1_balance_ok/CWRU_RMS_{}_{}.

2022-03-22 14:32:11 1240

原创 210316-针对类别不平衡数据集PyTorch实现每个Batch中出现所有类别及数量近似（待整理）

def prepare_dataloader(X, Y, P=None, dim='2D', batch_size=32, drop_last=True): from torch.utils.data import Dataset, DataLoader import torch import numpy as np from collections import Counter from torch.utils.data.sampler import Weighte

2022-03-16 14:19:21 1893

原创 220315-PyTorch中target为浮点数float时的交叉熵loss计算

Pytorch默认的交叉熵函数使用loss=(pred=浮点数, target=整数)的形式# Example of target with class indicesloss = nn.CrossEntropyLoss()input = torch.randn(3, 5, requires_grad=True)target = torch.empty(3, dtype=torch.long).random_(5)output = loss(input, target)output.back.

2022-03-15 14:03:55 1724

原创 220315-MultilinearMap的Numpy与Pytorch实现

import numpy as npimport torchdef multulinear_map_numpy(F, P): import numpy as np F1 = F[:,:,np.newaxis] P1 = P[:,:,np.newaxis] F2 = np.repeat(F1, P.shape[1], 2) P2 = np.repeat(P1, F.shape[1], 2).swapaxes(1,2) P3 = F2*P2 P4

2022-03-15 08:55:00 1389

原创 220309-Oh My Zsh “agnoster” theme 字体显示方框

Step 1:git clone https://github.com/abertsch/Menlo-for-Powerline.gitStep 2:将所有文件复制到字体库Step 3:设置VsCode中的Terminal fontReference: https://cloverinks.medium.com/oh-my-zsh-agnoster-theme-not-showing-correct-font-on-vscode-ubuntu-47b5e8dcbada...

2022-03-09 18:10:55 477

原创 220221-VsCode配置花生壳内网渗透

主要是HostName与Port的配置，此两项由花生壳提供Host homepcHostName [这里填写花生壳映射ip的地址，而非local的地址]User [这里填写local的用户名]Port [这里填写花生壳的宽口号，而非local的端口号]IdentityFile /Users/[这里填写loca用户名]/.ssh/id_rsa...

2022-02-21 13:18:48 306

原创 220218-Pandas保存Acc与Std并导出Excel

import numpy as npimport pandas as pd def save_acc_std_results(data_mean, data_std, methods, tasks): from styleframe import StyleFrame data_stat = np.zeros_like(data_mean, dtype='object') m, n = data_mean.shape for r in range(m):

2022-02-18 10:44:56 761

原创 220114-模型预测输出置信度的可视化

import pandas as pddf = pd.DataFrame(Pre.numpy())df['Cfd'] = np.max(Pre.numpy(), axis=1)df['Cfd'] = (df['Cfd'] - df['Cfd'].min())/(df['Cfd'].max() - df['Cfd'].min())df['Pre'] = np.argmax(Pre.numpy(), axis=1)df['Tar'] = target_Ydf['Tag'] = [1 if p el

2022-01-14 11:39:27 606

原创 220112-3步Word双栏排版图表单栏置顶居中

Step1: 插入-文本框-绘制文本框Step2: 复制图/表到新建的文本框内，调整到合适大小Step3: 右键文本框-更多布局选项-设置如下，最后确定即可实现标题效果

2022-01-12 09:23:43 1320

原创 211230-一维时间序列随机连续采样

import numpy as npdata = np.arange(200).reshape(-1,1)fea = 6sap = 10interval = 32ctn = 5zone = np.random.randint(1, len(data)-fea-interval*ctn, sap)print('Random sampling')print(data.shape)print(zone)print(len(zone),'\n')ctn_arr = []zone = n

2021-12-30 13:22:30 502

原创 211214-Word公式自动化AutoField

Word中6步公式居中及自动化编号更行https://guokliu.blog.csdn.net/Step 1: 插入公式Step 2: 定义格式Step 3: 插入域码Step 4: 回车显示Step 5: 重复二三Step 6: 全选更新

2021-12-14 10:34:39 866

原创 211213-LaTex生成Reference但无法显示Journal

问题：利用Zotero导出bib文件，然后LaTex排版，参考文献中没有期刊的名字。-分析Zotero生成的期刊名对应的字段是journaltitle，而LaTex在unsrt排序中，只识别journal，故此可将journaltitle=替换成journal=，问题解决。...

2021-12-13 14:22:31 948 1

原创 211212-LaTex排版IEEE双栏公式的3种方法

方法1: IEEE官方修改（本文推荐）\begin{figure*} \begin{equation} \label{eq-13} +++++++++++++++++ Put equation here +++++++++++++++++ \end{equation} \hrulefill % \vspace*{4pt}\end{figure*}方法2: IEEE官方，\begin{document}前插入\newcounter{my.

2021-12-12 18:57:45 3809

原创 211209-自定义Sklearn混淆矩阵及参数理解

在计算转移矩阵的时候，预测类别可能是实际类别的子集，sklearn自带的混淆矩阵不能直接使用，此处自定义，并进一步理解sklearn中混淆矩阵的定义。混淆矩阵Matlab中X-Target, Y-Output (Link)Sklearn中X-Output, Y-Target (Link)代码import numpy as npfrom sklearn.metrics import confusion_matrixdef my_confusion_matrix(y_true, y

2021-12-09 15:10:31 1212

原创 211201-LaTex插入图片作为公式且居中

Demo 1\begin{equation}[Son-Goku]\adjincludegraphics[valign=c, width=0.25in]{Figures/super1.png} + [Vegeta]\adjincludegraphics[valign=c, width=0.25in]{Figures/super2.png} = [Kugeta]\adjincludegraphics[valign=c, width=0.25in]{Figures/super3.png}\end{eq..

2021-12-02 13:12:48 380

原创 211130-Python谱图(Spectogram)分析Demo

STFT 算法对信号分段进行FFT处理，每次处理的结果都是谱图中的一列；每段芯好的长度越短，时间轴上的精度越高，频率轴上的精度越低时间轴和频率轴的分辨率是一堆不可调和的矛盾根据傅立叶变换的不确定原理，我们不能同时获得频率和时间的高分辨率specgram() 函数第一个参数：信号的组数第二个参数：FFT的长度第三个参数：信号的采样频率overlap：连续两块数据之间的重叠部分的长度，该参数越接近FFT的长度，FFT运算的次数越多，时间轴上的精度越大import numpy a

2021-11-30 10:53:04 1267 1

原创 211112-多维数据分布MMD相似性计算Demo

运行结果main函数import matplotlib.pyplot as pltimport numpy as npmean1 = [0, 0]cov1 = [[5, 0], [-5, 10]] # diagonal covariancemean2 = [10, 0]cov2 = [[5, 0], [5, 10]] # diagonal covariancex1, y1 = np.random.multivariate_normal(mean1, cov1, 500..

2021-11-12 14:13:24 1476

原创 211111-Python十进制转二十六(用于Excel表格自动化统计)

base = 'ABCDEFGHIJKLMNOPQRSTUVWXYZ'def decTo26(num): l = [] while True: num,rem = divmod(num, 26) print(num, rem) l.append(base[rem-1]) if num == 0: print(l) return ''.join(l[::-1])num = 27

2021-11-11 17:36:57 561

原创 211024-Pytorch模型训练中显示进度条

# ! with tqdm def fit(self, args, model, device, train_loader, optimizer, epoch, vis=True): # print('Epochs: ', epoch) correct = 0 L = len(train_loader.dataset) model.train() with tqdm(train_loader, unit="bat...

2021-10-24 11:13:04 19285

原创 211014-Python短时傅里叶变换

from scipy import signalimport matplotlib.pyplot as pltrng = np.random.default_rng()fs = 10e3N = 1e5amp = 2 * np.sqrt(2)noise_power = 0.01 * fs / 2time = np.arange(N) / float(fs)mod = 500*np.cos(2*np.pi*0.25*time)carrier = amp * np.sin(2*np.pi*3e

2021-10-14 18:59:34 157

matlab图形空白裁剪及子图调整

190313-1行命令裁剪Matlab图像空白及调整子图间距；

2019-03-13

SurfEasy-Chrome代理插件

SurfEasy-Chrome代理插件;SurfEasy-Chrome代理插件;SurfEasy-Chrome代理插件;

2018-04-22

recurrent neural network without a phd

recurrent neural network without a phd.pdf 出处：https://github.com/martin-gorner/tensorflow-rnn-shakespeare

2018-04-16

吴恩达-深度学习-第五课-第二周课件-截图版pdf版

吴恩达-深度学习-第五课-第二周课件-视频截图版pdf版（没找到对应的ppt版本），介意者请勿下载，欢迎分享ppt原版。

2018-04-14

windows10设置豆沙绿-双击执行即可

windows10设置豆沙绿-双击执行即可；windows10设置豆沙绿-双击执行即可

2018-03-29

A jupyter note result html file for Surface defect detection

2018-03-04

MNIST_data

MNIST_data手写体数字识别（友情提示-与py文件放在同一目录内）： import tensorflow.examples.tutorials.mnist.input_data as input_data mnist_data = input_data.read_data_sets('MNIST_data', one_hot=True)

2018-02-05

Chrome and SurfEasy Extension for Linux

Chrome and SurfEasy Extension for Linux。Chrome and SurfEasy Extension for Linux.

2017-12-02

斯坦福大学凸优化课程Convex Optimization字幕及Youtube字幕下载代码

http://blog.csdn.net/qq_33039859/article/details/78476993

2017-11-08

windows10桌面美化

window10桌面美化参见 http://blog.csdn.net/qq_33039859/article/details/78387485

2017-10-29

scipy-0.19.1-cp36-cp36m-win_amd64.whl

2017-07-16

a mov file with dog bark

2017-06-28

tensorflow model save and restore example

2017-06-15

tensorflow_gpu-1.2.0rc2-cp36-cp36m-win_amd64.whl

2017-06-14

tensorflow-1.2.0rc2-cp36-cp36m-win_amd64.whl

2017-06-14

deeplearning without a PHD (code print)

2017-06-12

pdf code pass

2017-06-12

deeplearning without a PHD

2017谷歌云大会，深度学习代码示例

2017-06-12

python ds evidence theory code

ds 证据理论 python 代码

2017-06-12

matlab ds evidence theory

ds 证据理论 matlab 代码

2017-06-12

《程序员的数学2-3》Ruby程序+Gnuplot+代码+勘误+win7补丁

2017-01-19

2017大数据发展趋势Top10白皮书

2016 was a landmark year for big data with more organizations storing, processing, and extracting value from data of all forms and sizes. In 2017, systems that support large volumes of both structured and unstructured data will continue to rise. The market will demand platforms that help data custod

2017-01-12

空空如也

TA创建的收藏夹 TA关注的收藏夹

TA关注的人