Implementation of various Machine Learning, Deep Learning methods. Some of the algorithms are coded from scratch, some are based on available libraries.

0 - Dimensionality Reduction Methods
Comparison between PCA, CCA, KPCA, t-SNE & ISOMAP.

1 - Regression methods (Implemented from scratch)
A. Linear Regression to predict House Prices.
B. Logsitic Regression (Stochastic Gradient Descent) to predict Employement.

2 - Deep Learning
A. Implement Deep Neural Nets (TensorFlow) with RELU activation units to classify MNIST data.
B. Convolutional Neural Nets (TensorFlow) to classify FASHION MNIST data.
C. ARIMA vs. LSTM (Keras) for BitCoin Price prediciton.
C. Deep Ladder Nets (Neural Nets modelled in TensorFlow) to classify MNIST data, given only 1% of the data is labelled. Inspired by https://github.com/CuriousAI/ladder

3 - Twitter Language Model
Predict Language using a Character Level Unigram Markov Model.

4 - Decision Trees (Implemented from scratch)
A. AdaBoost: Coding of algo to predict probability of Breast Cancer.
B. CART: Coding of algo on IRIS dataset.

5 - Ensemble Learning
Prediciton of Survival on TITANIC dataset. The final prediciton was generated using an ensemble of Random Forest, GradBoost, SVM, LogReg, AdaBoost learners that were finally passed through XGBoost.

6 - Expectation Maximization

7 - Generative Adverserial Network:
Implementation of a GAN with fully-connected layers for both Generator and Discriminator to learn a multimodal distribution. The loss used for training was Wasserstein Loss. Inspired by https://github.com/igul222/improved_wgan_training