amalik's repositories
categorical-feature-importances-without-one-hot-encoding-dummies
Feature Importance of categorical variables by converting them into dummy variables (One-hot-encoding) can skewed or hard to interpret results. Here I present a method to get around this problem using H2O.
Audio-driven-TalkingFace-HeadPose
Code for "Audio-driven Talking Face Video Generation with Learning-based Personalized Head Pose"
breast-cancer-detection-with-SVM
Breast cancer predictions using UCI's Breast cancer Wisconsin dataset
first-order-model
This repository contains the source code for the paper First Order Motion Model for Image Animation
galaxy-zoo
Code to perform morphological classifications of galaxies using machine learning
halo_conc-regression-ML
In this code, we will carry out a simple regression task. For this we will first download a public dark matter halo catalogue from the Bolshoi simulation. We will then use the Pandas library to analyse this halo catalogue and to identify correlation between different halo properties. In the next step, we will use the Scikit-Learn library to predict the halo concentration from the other halo properties. For this we will test simple linear regression, a decision tree, and random forests. Finally, we will determine which regression algorithm performs best with respect to the test data set.
kbroman.github.io
Karl Broman's website
MCMC-cluster-mass
Code to estimate the mass of a galaxy cluster with weak lensing using Monte Carlo Markov Chains (MCMC).
Misc-demostrations-using-machine-learning
This section includes miscellaneous simple demonstrations using machine learning on various different kind of common problems.
models_K2
Models built with TensorFlow
Photometric_Z-Regression
In this notebook we used decision tree and random forests algorithms to learn a mapping between galaxy magnitudes and colors and the spectroscopic redshift. We also constructed some learning curves to determine the best choices of the algorithms. We then used feature importance to understand which of the features give us the most predictive power, and then train a very simile algorithm with reduced feature list.
predicting-term-deposit-subscription-tendencies-of-clients
Here we will predict whether a client is likely to subscribe to a term deposit and will identify the features that play an important role in determining this.