README.md

STATISTICAL-CONCEPTS

Repo of abstract explaining the statistical concepts behind Machine learning algorithms and their python implementation.

---

Data Transformation | .pdf | code

Data transformation is a crucial part of machine learning since an uncurated data would affect the performance of a machine learning algorithm (model). Emphasis is placed on the significance of transformation for better learning.

---

Regression | .pdf | code

Regression models make use of input data attributes or features (explanatory or independent variables) and their corresponding continuous numeric output values (dependent or outcome variable) to learn specific relationships and associations between the inputs and their corresponding outputs.

---

Classification | .pdf | code

Classification is a supervised machine learning approach to categorizing data into distinct number of classes where we can assign label to each class

---

Model Validation | .pdf | code

Model validation is an essential and perhaps the most important of machine learning process. This is because it helps in checking the stability of the model and stops it from overfitting or underfitting on training datasets.

---

Evaluation Metrics | .pdf | code

Model/hypothsesis evaluation is a machine learning approach for assessing the performance of an algorithm on a dataset. The objective is to reduce the error between actual data and prediction.

---

Activation and Cost functions | code

Activation and Cost functions used in MAchine/Deep Learning models inlcuding Sigmoid, Relu (Rectified Linear Unit), LeakyRelu, Hinge loss.

---

Kernel functions | code

Module containing frequently used machine learning kernels. Including multiple kernel methods used in our recently concluded paper.

---

Kernel Methods | pdf

Repo of advance machine learning methods based on kernel learning. Implement kernel methods for Principal Component Analysis (PCA), Kmeans, Logistic Regression and Support Vector Data Description.

---

Variational Autoencoder (VAE) | pdf

Mathematical details behind Variational AutoEncoder method and a derivation of Evidence lower bounds (ELBO) for Gaussian distributions and Inverse-Wishart distributions.

---