studyvcmfc的专栏

指数滑动平均法_百度百科。

将数据上传到 S3在上个 notebook 中，你应该使用给定剽窃/非剽窃文本数据语料库的特征和类别标签创建了两个文件：training.csv和test.csv文件。以下单元格将加载一些 AWS SageMaker 库并创建一个默认存储桶。创建此存储桶后，你可以将本地存储的数据上传到 S3。将训练和测试.csv特征文件保存到本地。你可以在 SageMaker 中运行第二个 notebook“2_Plagiarism_Feature_Engineering”，或者使用 Jupyte...

Quora Question Pairs 思路记录https://blog.youkuaiyun.com/u010863933/article/details/79570902?ops_request_misc=%257B%2522request%255Fid%2522%253A%2522159713848619724845042224%2522%252C%2522scm%2522%253A%252220140713.130102334..%2522%257D&request_id=159713848619

SAGEMAKER上部署需要引入sagemakder的API from sagemaker.pytorch import PyTorchModel

报错 RuntimeError: size mismatch, m1: [10 x 3], m2: [2 x 10]原因：train.csv 为100x4 4列 第一列 标签是否剽窃 后三列为取的特征值输入应该为3，结果定义模型的时候，选择的input_features =2 导致数据结构不匹配引起from sagemaker.pytorch import PyTorch# your import and estimator code, hereoutput_path=...

https://www.cnblogs.com/zwt20120701/p/10851081.html

Persistent storage volumes for your data using Amazon Elastic Block Store (Amazon EBS), known asAmazon EBS volumeshttps://docs.aws.amazon.com/AWSEC2/latest/UserGuide/concepts.html

What is Amazon EC2?PDFKindleRSSAmazon Elastic Compute Cloud (Amazon EC2) provides scalable computing capacity in the Amazon Web Services (AWS) cloud. Using Amazon EC2 eliminates your need to invest in hardware up front, so you can develop and deplo

https://stats.stackexchange.com/questions/2691/making-sense-of-principal-component-analysis-eigenvectors-eigenvalues

一开始由于我本身软件开发的基础，短期内很快就完成了前端开发纳米学位的学习。但这招“闪电战”在机器学习面前立马不管用了。“强化学习”和“深度学习”这些课题都很难，大大超出了我目前的能力范围，我卡了很久。最后决定看不懂的暂时跳过，先坚持自己的计划，把纳米学位修完，然后再慢慢消化和复习学过的内容。https://zhuanlan.zhihu.com/p/45502088...

# Freeze parameters so we don't backprop through them#print("model.parameters()=",type(model.parameters()))import torchfrom torch import nnfrom torch import optimimport torch.nn.functional as Ffrom torchvision import datasets, transforms, models%ma

https://zengwenqi.blog.youkuaiyun.com/article/details/96282788

# Build a feed-forward networkmodel = nn.Sequential(nn.Linear(784, 128), nn.ReLU(), nn.Linear(128, 64), nn.ReLU(), nn.Linear(64, 10))# Define the losscriterion = ...

def softmax(x): ## TODO: Implement the softmax function here #print("torch.exp(x)=",torch.exp(x)) s = torch.sum(torch.exp(x),dim=1) print("s.size=",s.size()) print("s.view(-1,1)=",s.view(-1,1).size()) #有问题的return （torch.exp(x)/s).vi...

https://classroom.udacity.com/nanodegrees/nd009-cn-advanced/parts/5f4d630c-d15a-412c-aaeb-b57ad61cd03c/modules/3aa9e812-62cd-4ae3-8fc4-593538f08455/lessons/9b014a97-2267-4f1b-af97-284b7dac2a58/concepts/7921d0d6-b19c-474e-b8fd-2379c69d75ed

https://www.cnblogs.com/paulonetwo/p/10078961.html

http://yann.lecun.com/exdb/mnist/

向量入门https://www.khanacademy.org/math/linear-algebra/vectors-and-spaces/vectors/v/vector-introduction-linear-algebra矩阵入门https://www.khanacademy.org/math/precalculus-2018/precalc-matrices

keras如何解决局部最小问题https://ruder.io/optimizing-gradient-descent/index.html#rmspropkeras如何解决局部最小问题 APIhttps://keras.io/api/optimizers/

从EPOCH20开始出现了过拟合曲线，表征是此时测试集的误差达到了最低，再往后的训练导致训练集的误差虽然进一步减小，但是测试集合的误差却持续走高，这就是过拟合的明显特征https://classroom.udacity.com/nanodegrees/nd009-cn-advanced/parts/5f4d630c-d15a-412c-aaeb-b57ad61cd03c/modules/3aa9e812-62cd-4ae3-8fc4-593538f08455/lessons/2f907351-...

构建一个简单的多层前向反馈神经网络以解决 XOR 问题。将第一层设为Dense()层，并将节点数设为8，且input_dim设为 2。 在第二层之后使用 softmax 激活函数。 将输出层节点设为 2，因为输出只有 2 个类别。 在输出层之后使用 softmax 激活函数。 对模型运行 10 个 epoch。...

https://keras.io/https://www.tensorflow.org/http://caffe.berkeleyvision.org/http://deeplearning.net/software/theano/https://scikit-learn.org/stable/

https://www.jianshu.com/p/aee95376d8b8

https://blog.youkuaiyun.com/gaotihong/article/details/100815858

https://blog.youkuaiyun.com/grape875499765/article/details/78635931?utm_medium=distribute.pc_relevant.none-task-blog-OPENSEARCH-16&depth_1-utm_source=distribute.pc_relevant.none-task-blog-OPENSEARCH-16

https://numpy.org/devdocs/reference/generated/numpy.where.html?highlight=where#numpy.where

https://www.jianshu.com/p/eb778a3ef633

https://scikit-learn.org/stable/auto_examples/model_selection/plot_learning_curve.html

在所有我们预测的阳性结果中，有多少是正确的？ 精度 = 真阳性 / （真阳性 + 假阳性）

Turns Out…We can see from the scores above that our Naive Bayes model actually does a pretty good job of classifying spam and “ham.” However, let’s take a look at a few additional models to see if w...

边际误差计算在本部分，我们将会计算 SVM 中两个间隔之间的距离。首先，W = (w_1, w_2)W=(w1​,w2​)，x = (x_1,x_2)x=(x1​,x2​)，并且Wx = w_1x_1 + w_2x_2Wx=w1​x1​+w2​x2​.请注意，在这里我们有三条线，方程如下：Wx+b=1Wx+b=1 Wx+b=0Wx+b=0 Wx+b=-1Wx+b=−1由于...

https://www.cnblogs.com/pinard/p/6074222.html

机器学习数据集库https://archive.ics.uci.edu/ml/datasets/SMS+Spam+Collectionhttps://archive.ics.uci.edu/ml/machine-learning-databases/00228/

后验概率的本质：是根据先验概率所描述的现象反推该现象来自先验概率的概率，即用修正一词来表达比较合适

算法调整原理 如果点分类正确，则什么也不做。 如果点分类为正，但是标签为负，则分别减去αp,αq,和α至w_1, w_2,w1​,w2​,和bb 如果点分类为负，但是标签为正，则分别将αp,αq,和 α加到w_1, w_2,w1​,w2​,和bb上感知器算法掌握了感知器技巧后，我们就可以编写完整的感知器运算的算法了！下面的视频将介绍感知器算法的伪代码...

http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Lasso.html

#HSV 8*8#hsv[[[ 75 2 236] [165 4 238] [171 25 206] [171 36 196] [173 23 198] [170 10 227] [ 80 3 237] [ 0 0 238]][[ 50 3 237] [ 20 3 223] [171 91 204] [171 1...

https://blog.youkuaiyun.com/ColdWindHA/article/details/82080176

https://www.bilibili.com/video/av62245681?from=search&seid=5806181273848770603

https://machinelearningmastery.com/how-to-one-hot-encode-sequence-data-in-python/