This tutorial shows how to use imperative tensor layer of Alea TK for general purpose computations. We program a Monte Carlo simulation to approximate \(\pi\).

MNIST is a computer vision dataset consisting of handwritten digits. It is a good dataset to try out machine learning techniques and pattern recognition methods on real-world data. This tutorial shows how to train a multinomial logistic regression, a multilayer perceptron and a convolutional neural network to predict the digit from an image.

This tutorial performs word-level prediction using a 2 layer LSTM model trained on the Penn Treebank (PTB) dataset.