This Repo consists of Machine Learning based projects that are build on opensource frameworks like scikit-learn, keras , tensorflow
.
Folder from_stratch
consists of ML projects that are build from scratch .
-
Neural Networks from scratch ,
This project was build with the intent of gaining more insighst about neural networks , Stochastic-gradient-descent , Logistic regresssion
-
Kmeans ,
Mimicks the scikit-learn's kmeans clustering algorithm from scratch using just numpy.
-
Implementation of gradient-descent without using any ML frameworks , just using
numpy
for computation andmatplotlib
for visualization. -
Implementation of Newtons-method ( Second-order-Optimisation ) from scratch & using
numpy
for computation andmatplotlib
for visualization.
Folder supervised
consists of Real-time-applications / Projects based on supervised Learning.
-
This provides some insights about comparing the classifiers , basic hyperparameters for a particular dataset.
-
A Realtime twitter-api which fetches tweets based on topic , performs sentiment analysis.
-
Time-series based problem , build using keras api .
-
With grouplens dataset , providing collaborative + content based recommendations.
-
Using tensorflow's convnet to perform deep dream , gradient-descent on images , video (under-development).
-
Recognising whether the given image is humane or not ,using convnet built on keras.
-
genetic_programming, (under-developement)
Using genetic programming to find the hyperparameters for a given dataset.
-
emnist, (under-developement)
Project to recognise handwritten alphanumeric character with dataset using emnist (extention of mnist).
Folder unsupervised
consists of Real-time-usecases / Projects based on unsupervised Learning.
-
Converting Images of one color system (RGB) to another color system (Grey) using
Autoencoders
. -
Finding the dorminant colors present in the picture , frequency of occurance.
-
wordvec, (under-developement)
A simple service that lets you find related words that is related to "Arts" industry.
Folder datamining
consists of data-mining based projects.
-
orthonearme, (under-development)
Web-app that lets you find the Ortho specialist near your proximity , based on rating / comments.
Credits:
Thanks to siraj raval
.