Skip to content

hottbox/hottbox-tutorials

Repository files navigation

HOTTBOX tutorials

Binder

This repository contains a series of tutorials on how to use hottbox.

Local Installation

In order to get started you need to clone this repository and install packages specified in requirements.txt:

git clone https://github.com/hottbox/hottbox-tutorials

cd hottbox-tutorials

pip install -r requirements.txt

If you are on Unix and have anaconda installed, you can execute bootstrap_venv.sh. This script will prepare a new virtual environment for these tutorials.:

git clone https://github.com/hottbox/hottbox-tutorials

source bootstrap_venv.sh

Table of contents:

Focus of the tutorial Static notebook on github.com Interactive notebook on mybinder.org
  1. N-dimensional arrays and its functionality: Tensor
Tutorial1 ti1
  1. Efficient representation of N-dimensional arrays: TensorCPD, TensorTKD, TensorTT
Tutorial2 ti2
  1. Fundamental tensor decompositions: CPD, HOSVD, HOOI, TTSVD
Tutorial3 ti3
  1. Ecosystem of Tensor class and transformations
Tutorial4 ti4
  1. Tensor meta information and pandas integration
Tutorial5 ti5

Data used in these tutorials

All data for these tutorials can be found under data/ directory.

Short description of datasets

  • ETH80 dataset

    This dataset consists of 3,280 images of natural objects from 8 categories (apple, car, cow, cup, dog, horse, pera, tomato), each containing 10 objects with 41 views per object. More info about this dataset can be found on here.

Short description of files with data

  1. data/ETH80/basic_066-063.npy

    Contains only one RGB image of one object from each category, which makes it a total of 8 samples. The view point identifier for all of them is 066-063. These images are 128 by 128 pixes and are stored in the unfolded form. Thus, when this file is read by numpy it outputs array with 8 rows and 128*128*3 = 49152 columns.