Basic Setup of environment and enrollment in Udacity's Machine Learning intro course and completed Introduction videos.
The course is amazing and excited about 100 days of ML Code.
Just enrolled in this course: https://in.udacity.com/course/intro-to-machine-learning--ud120-india
Learnt how exactly gradient descent works with mathematical derivations and also learnt Naive bayes for classification and implemented using sklearn
Gone crazy in mathematical proofs for gradient descent but very important to understand why we are doing everything and next sklearn is very easy to use.
Learnt and solved problems on Bayes rule and learnt it's applications. Started developing the mini project on naive bayes.
Once J.K.Rowling published a book name under false author name but the computers were able to identify that it was her and it's exciting to know that it can be done using naive bayes.
Learnt about McCulloh pits neuron, their advantages and disadvantages and moved on to perceptron and also learnt perceptron algorithm
Thought that I could the complete project started on day:2 but I couldn't as the dataset was huge, so learnt moved onto deep learning and learnt the perceptron stuff.
Set up the python 2.7 environment ans finally completed the mini project on naive bayes and acheived an accuracy of 97.32%
The dataset was huge and it took much time to download and extract but it is good to do some mini projects as they clear the concepts.
Learnt about linearly saperable boolean functions, Sigmoid neuron, the typical machine learning setup.
Understood the need for using sigmoid neuron, and the basic machine learning things needed to solve any problem.
Learnt about feed forward forward neural networks and how to learn the parameters the feed forward neural networks.
It's interesting to learn how to apply gradient descent for multilayer neural networks and the proofs.
Learnt about Entropy, Cross Entropy and their derivations and understood which loss function to choose depending on outputs.
The deeper you go into deep learning the more it is confusing. It took so much time to understand the difference between entropy and cross entropy.
Learnt the backpropagation intuition and the derivations of the gradent with respect to the output units.
Really exciting to know how exaclty backpropagation works with mathematical derivations, but just completed on output layer need to still calculate with respect to hidden layers, weights and biases
Started learning SVM's in Udacity's Intro to Machine Learning Course..
I really have a lot of confusion with SVM's, need to understand them better.
Learnt about linear SVM and coded using SKlearn.
The linear svm is easy to understand but need to know about several parameters used in it.
Completed learning SVM including non-linear, and also learnt about gamma and C parameter.
Interesting to learn about various kernels, started doing a project on SVM.
Successfully completed project using SVM, the same project done using naive bayes but svm gives an accuray more than 99%.
SVM's really take a lot of time in training but ofcourse gave a good accuracy.
Started learning Decision trees from Udacity's Intro to Machine Learning course, learnt about basic decision tree and coded using skklearn.
Decision trees are more prone to overfitting, need to learn if there is any way to reduce that.
Learnt about Entropy, Information Gain, Bias and Variance. Started building the project on Decision Trees
Understood decision trees very clearly, they are just simple like if else blocks.
Successfully completed the project on Decision trees(the same project done using naive bayes) and got an accuracy of 97%
Decision trees might be overfitting if we have more features, but in general gives good results
Intersted in developing a project on Breast cancer prediction, so started learning Flask to build backend
Flask is very easy to learn and yet powerful, learnt basics quickly
Started working on Breast cancer prediction, collected the data and done some preprocessing
Lucky that most of the data is good, so made it easy to process
https://www.kaggle.com/uciml/breast-cancer-wisconsin-data
Built complte Prediction using SVM and also built an API to make the requests to server
Too complex in this step, tried with ANN but it is over fitting the model so used SVM
Successfully completed the project including the frontend with bootstrap, working great and yields an accuracy of more than 98%
It was really a wonderful experience to build the complete project. so happy
https://github.com/SurajChinna/Breast-Cancer-Prediction
Started building a project that could predict the relationship between two sentences of a paragraph and returns the percentage using wordnet
I really know nothing about nltk and wordnet but still trying to build a project
Woo! Successfully completed the project but that is not accurate but doing well
Felt really tough to jump into project without knowing anythig but somehow managed to complete.
https://github.com/SurajChinna/Breast-Cancer-Prediction
Learnt the basic concepts of KNN and how it works
KNN is pretty easy just some math.
Back into Udacity's Intro to Machine Learning course, started lesson datasets and learning about enron
Enron was a huge organisation that went bankrupt and it is interesting to study the reasons of failure
Just analysed the data-set of enron and various features
Finally, done with enron and need to start Regression
Started the lesson regression.
Just looked how and where the regression is used.
Learnt in greater detail about regression, various evaluation metrics and differences between regression and classification.
Regression is pretty easy but learnt only linear regression
Learnt evaluation metrics in Regression, and completed a small project that could fit the line between age and salary.
Project was good but there were some outliers that disturbed the line so need to learn about outliers
Learnt some basic stuff about outliers and a simple algo to remove them.
Outliers are really disturbing..need to do some project
Learnt and Written a Small practical code to identify and also remove outliers.
It's a good experience to check and remove the outliers.
Understood the KMeans algorithm for clustering and its limitations.
KMeans is really basic but it makes a lot of mistakes due to initialisiations of centroids.
Got more idea about KMeans and completed the mini project in clustering.
need to learn feature scaling.
Started Feature Scaling lesson in Udacity's Intro to Machine Learning course. Learnt what is feature scaling and it's implementation in sklearn
Found feature scaling as a very important step for some algorithms like SVM and KMeans clustering.
Learnt when to apply feature scaling and completed the mini project on feature scaling
Feature scaling can be confusing but pretty interesting.
Started Text Learning lesson in Udacity and learnt about bag of words and how to represent them in sklearn
It's really hard to deal with language and a lot of confusion. I have to dig deeper
Learnt about stemmers and also learnt how to use them in sklearn and implemented SnowballStemmer
feeling confident as going more into text learning
Learnt about tfidf representation and saw how to implement in sklearn and starting a project in text learning
Need to apply all learnt concepts in project
Completed the project successfully that takes a email and does the preprocessing
Need to learn more in nltk. so wanted to start nltk tutorials on youtube
Learnt about word tokenizers, sentence tokenizers and stop words in nltk
Got interested in nltk, so started youtube videos from sentdex. They are really cool!!
Learnt about Stemming and used Porter stemmer and next learnt part of speech tagging
Getting so much interest in NLP, with the concepts learnt will do a project on creating a dictionary and part of speech tagging using PyQt
Learnt about Chunking and Chinking and some regular expressions and done name phrase chunking on some text
It's really confusing to learn chunking, took too much time, and hard to understand regular expressions. But anyway understood some good concepts
Studied a lot about Named Entity recognition, implemented that, and read about IOB representations
Named Entity Recognition is very important but sometimes it misclassifies the tags. may be I should learn spaCy in some free time
Understood how to use different corpora and learnt about lemmatizing
Lemmatizing looks better than stemming and working with huge corpora is good
Learnt about wordnet, to get synonyms and antonyms of word, to get definitions and examples and also computing the similarity of two words using wordnet
Wordnet is huge and got confused between synsets and lemmas but very useful in nlp and we can do a lot with wordnet
Started doing the project of movie review classification whether positive or negative. completed some portion of importing and storing in lists
Excited in doing projects as most of the learning happens there
Completed the projet of movie review classification using NaiveBayes and pickled the classifier. Getting an accuracy of nearly 80
Now feeling that I have to do more and more projects in NLP
Started doing real world project on movie review classification and got data from IMDB. just cleaned the data and arranged
Wanted to do this project more user friendly and easy to use.
Formatted data according to required format and removed the useless words and cleaned the data.
The most toughest job is data cleaning and successfully completed that. next need to train and test on data
Revised and learnt some concepts in numpy from youtube videos.
Text classification is under process and without wasting time I wanted to start another machine learning tutorial.
Started learning pandas from here. Completed learning DataFrame basics, ways of creating data frames and different functions to read and write csv and excel files
As pandas is very important in data analysis, started learning it in depth.
Learnt how to handle missing data and various methods to handle them, to use replace function to replace special characters
Now feeling confident as I have completed handling missing data as it is the most important in data munging but still have so much to study in pandas.
Learnt how to handle missing data using sklearn's Imputer, and also how to encode categorical data using sklearn
Getting into Data preprocessing so that I can use these in future machine learning models and still have so much to study in Data Preprocessing
Studied about feature scaling like standardisation and normalization and learnt how to implement them in sklearn and also learnt how to split the test and train data
Completed most of the steps in data preprocessing and got ready for starting machine learning topics
Learnt the Simple Linear Regression and also completed the project predcting the salaries based on years of experience
Super excited to start ML from basics, and completed the project, need to learn about more Regression Models
Learnt about Multiple Linear Regression, Homoscedasticity, Multicollinearity and Dummy variable trap
studied a lot of theory regarding MLR, now need to start a project
Learnt about methods of building a model, like backward eliminaton, forward selection, bidirectional elimination
Took so much time to understand, as these concepts are very complex.
Successfully completed the project on Multiple Linear Regression, that could predict the profit of a start up based on the amount they have spent on marketing, R&D etc.
Project is working fine, but need to use backward elimination to eliminate the preditors that are not very important
Hurray!! Used backward elimination to eliminate the features that are statistically less significant. Now predicted results are more closer to the actual results
Backward elimination was very confusing, but after watching lot of tutorials finally got good understanding of that. and yeah successfully completed the project.
Learnt Polynomial Regression and learnt how to use it with linear regression, completed the project on predicting the salary of employee based on their level using Polynomial Regression
Polynomial regression is just slight extention of multiple linear regression but quite useful as real world data is not linear.
Studied about Support Vector Regression and completed the project that could predict the salary of employee based their level
SVR was very easy to understand and implement but a bit confusing in gamma and C parameters
Studied about Decision tree regression and also completed the project that could predict the salary of employee based on their position level using Decision Tree Regression
Decision tree is not giving very accurate results and hence need to learn some ensemble methods
Studied about Random forest regression regression and also completed the project that could predict the salary of employee based on their position level using Random Forest Regression
Hmm..The accuracy has increased compared to Normal Decision trees and is very robust.
Learnt about Logistic Regression and the mathematical changes from linear regression and also the evaluation metrics for classification problems.
Moving from Regression to Classification, started logistic regression and completed the theory and next will do a project.
Successfully completed the project that classifies the users whether they will buy SUV based on their age and salary, gives and accuracy of 89%
Doing projects is more interesting, and next going to study about KNN classifier
Learnt about knn classifier and completed the project same project as logistic regression but got an accuracy of 93%
KNN got more accuracy because it is non linear model, and hence quite useful.
Completed learning about linear SVM and also done the same project as logistic regression and got an accuracy of 90%
Accuracy was quite low because I have used linear SVM, will study about various kernels and try to improve that
Completed the project on diabetes classification using KNN, but the model is giving low accuracy and needs to handle data carefully
Handling the data and trying to build with other models too
YaY!! Successfully completed the project and working great with an accuracy of nearly 81% using Logistic Regression
Data had no null values but has zero values so handled them and used Logistic Regression which is giving good accuracy
Learnt about various kernels in SVM like rbf, sigmoid and polynomial. Understood the kernel trick and how non linear classification is done
Understood the concepts and now improving the project that was done using linear SVM
Included the kernel SVM file for the previous repo. kernel SVM with rbf kernel gives great accuracy of 93%
The kernel SVM works great but is really broad and complex. confused with various kernels but understood basic kernels
Studied the intuition behing Naive Bayes algorithm and solved some problems using the bayes formula.
Now clearly understood the working of Naive Bayes and next going to code that out.
Completed the project of classification using Naive Bayes and got an accuracy of 91%.
Naive bayes is naive but it really works great for the classification using social media ads project.
Understood the intuition beging decision tree classifier and also completed the project of classification and got an accuracy of 91%
Decision trees are too old but works good and gives good accuracy but will imporove the efficiency using ensemble methods
Learnt about ensemble learning, Random forest classifier and also completed the same project of classification using Random Forest Classifier.
Hmm.. Random forest does a good job and gives an accuracy of 92% and next need to learn about improving the classification models