Originally published at https://dataaspirant.com on November 9, 2020.
Creating deep learning or machine learning models in local systems is like a cakewalk. Things get complicated when we try to replicate the same project setup in the cloud.
The two popular options we as a data science community have for managing project environments are anaconda environment and python virtualenv.
Which one did you use?
Else let me put the straight question which project environment is best for deploying data science projects in the cloud?
Anaconda? or python virtualenv environment?
Confused to answer this question, don’t blame your mind, just relax and read this article. You will get clear answers with many reasons or features to keep in mind while selecting the environment for your next data science projects. …
If you know the basics of deep learning, you might be aware of the information flow from one layer to the other layer. Information is passing from layer 1 nodes to the layer 2 nodes likewise. But how about information is flowing in the layer 1 nodes itself. This is where recurrent neural network, in short, RNN architecture, came out.
Suppose we are building a model to predict the next coming word. How do you do that?
In this case, we need the previous word information of the prior state/node along with the input at the current layer node to generate the next coming word.
Originally published at https://dataaspirant.com on November 5, 2020.
While the world is moving towards touchless technologies, computer vision algorithms plays a key role in building such applications to reduce human interventions.
One such application is recognizing human faces. Without knowing, we all are an integral part of face recognition application in our daily lives, ranging from unlocking mobile with face recognitions, google photos grouping, facial recognition in surveillance, pay by face techniques, and many more.
This article will explain how to build a face recognition application that detects the faces and counts the faces based on gender using OpenCV.
While checking the performance of regression models, the fundamental methods are r-squared and adjusted r-squared.
This question is ubiquitous in data scientist interviews too. The adjusted r-squared has an added advantage over the r-squared.
So learn the key differences between the r-squared and adjusted r-squared with advertising vs sales growth case study.
As a fresher, it’s very hard to get the data scientist job. The same applies to the people who are changing from other domains or technology to the data science field.
But if we follow a strategy to prepare to learn the data science field’s required skill set. We can undoubtedly get the first job as a data scientist.
This article helps you achieve this by providing 6 stage strategy to get your entry-level job as a data scientist.
In the article, we provided completely free resources (zero cost) to become a data scientist. In other words, the article’s intention shows how you can use the strategy and the free resources to become a data scientist.
If we miss any free and valuable resource, do let us know, we will include those in the article.
Activation functions are the building blocks for simple neural networks and the complex deep learning network architectures.
People tend to use these activation functions like try and error type while building deep learning networks.
But if you learn the properties of these activation function, you will know which activation function has to use where.
The article listed the popular activation functions used in the industry. If we miss any, do let us know, we would love to include that in the article.
In the banking domain, predicting the credit card fraudulent activities it the key pressing issue.
With the increase in technology, the way fraud is happing is also changed into new ways.
So these actives need to detect with machine learning or deep learning algorithms.
This article will build the famous credit card fraud detection model using the Kaggle credit card data.
We will also learn how to handle the imbalance problems.
With the availability of various deep learning frameworks, people tend to learn the deep learning concepts by using these frameworks as starting points. Still, it is worth knowing the building block concepts to build neural networks.
This article explains the basic concept with mathematical details to build neural networks; the knowledge gained will help understand the various deep learning models architectures in the long run.
The real word text is a combination of a high amount of noise and minimal insights.
So we need to apply different text preprocessing techniques before going to build models. If you think we can use a bunch of machine learning preprocessing techniques. Then we are on very very wrong track.
The natural language preprocessing techniques are entirely different from machine learning and deep learning data preprocessing techniques.
The below link explains the most popular 20+ text preprocessing techniques along with implementation in python.
If we miss any popular methods? Do let us know; we would love to know that.
Instead of building only one machine learning model to predict the future, how about building multiple machine learning models to predict the future?
How awesome it will be.
This is the main idea behind the ensemble methods. Bagging and Boosting are the most popular ensemble methods. Even though both bagging and boosting came from the same family called ensemble methods, they are still different.
So let us learn the significant difference between bagging and boosting methods.