段颖凡的博客 | Yingfan Blog

Data Analysis: 因果推断

因果推断方法介绍

Correlation Does Not Imply Causation 相关性通常是对称的，因果性通常是不对称的（单向箭头），相关性不一定说明了因果性，但因果性一般都会在统计层面导致相关性。这篇论文中提到了 “The Three Layer Causal Hierarchy”的概念。 level typical activity e...

Posted by Yingfan on June 10, 2022

Machine Learning: NLP with RNN and Attention

Char-RNN, Sentiment analysis, Bidirectional RNN, Attention

Posted by Yingfan on June 3, 2022

Machine Learning: RNN

RNN Introduction, LSTM, GRU, WaveNet

In this lecture we shall discuss how one can predict the future using recurrent neural networks (RNN). RNN Introduction RNNs are capable of handing sequential data. Sequential Data refers to any ...

Posted by Yingfan on May 23, 2022

Machine Learning: CNN

CNN introduction and Common Architectures

Convolutional neural networks (CNNs) are not restricted to visual perception: they are also successful at many other tasks, such as voice recognition and natural language processing. This blog focu...

Posted by Yingfan on May 16, 2022

Machine Learning: Training DNN

Initialization, optimization, transfer learning

Problems with training DNN Problems Training deeper networks consisting of 10s or 100s layers runs into the following problems: Vanishing gradient problem gradients get smaller and...

Posted by Yingfan on May 11, 2022

Machine Learning: Artificial Neural Networks

Artificial Neural Network with Keras

Artificial Neural Networks (ANN) is a Machine Learning model inspired by the networks of biological neurons found in our brains The Perceptron one of the simplest ANN architectures Threshold...

Posted by Yingfan on May 4, 2022

Machine Learning: 知识点汇总

面试各类问题知识点整理

1 基本概念 1.1 理解局部最优和全局最优优化问题一般分为局部最优和全局最优。（1）局部最优，就是在函数值空间的一个有限区域内寻找最小值；而全局最优，是在函数值空间整个区域寻找最小值问题（2）函数局部最小点是它的函数值小于或等于附近点的点，但是有可能大于较远距离的点（3）全局最小点是那种它的函数值小于或等于所有的可行点 1.2 各种常见算法日常使用机器学习的任务中，...

Posted by Yingfan on May 4, 2022

Machine Learning: Interview Questions

重点机器学习相关面试问题

线性回归如何判断是否符合正态分布？如果不符合正态分布要怎么办？计算数据分布的偏度和峰度值偏度是一种用来表示数据分布非对称程度的指标，为样本的三阶标准化矩，根据尾部可以判断是左偏（负）还是右偏（正）；峰度是描述总体中所有取值分布形态陡缓程度的指标；如果，偏度=0是为正态，峰度=...

Posted by Yingfan on April 30, 2022

Machine Learning: Unsupervised Learning

Introduction to unsupervised learning models

Unfinished Clustering K-Means Algorithm place k centroids randomly label other instances by assigning them to the cluster of the closest centroid update centroids...

Posted by Yingfan on April 27, 2022

Machine Learning: Dimensionality Reduction Techniques

Introduction to dimensionality reduction and python implementation

Importance The curse of the dimensionality When Machine Learning problems involve thousands or even millions of features for each training instance all these features make trai...