Generating digits by interpolating latent space with VAE

Version 1.4 (6.77 MB) by Kenta
Continuous digit generation using variational auto encoder (VAE) by interpolating the latent space. VAEによる数字の自動生成

114 Downloads

Updated 8 May 2020

View License

[English]
This demo generates a hand-written number gradually changing from a certail digit to other digits using variational auto encoder (VAE). The official documentation entitled "Train Variational Autoencoder (VAE) to Generate Images" was reffered for this demo as shown below.
(https://jp.mathworks.com/help/deeplearning/examples/train-a-variational-autoencoder-vae-to-generate-images.html)
The original paper is shown below
Kingma, Diederik P and Welling, Max. Auto-Encoding Variational Bayes. In The 2nd International Conference on Learning Representations (ICLR), 2013.
Note that this kind of demo has been already made, for example,
http://dpkingma.com/sgvb_mnist_demo/demo.html.
https://ml-showcase.paperspace.com/projects/variational-autoencoder-vae
https://qiita.com/kenmatsu4/items/b029d697e9995d93aa24
Here in this demo, the code with MATLAB is shown.
For example, VAE can be used for anomaly detection.
https://jp.mathworks.com/matlabcentral/fileexchange/73283-anomaly-detection-using-variational-autoencoder-vae?s_tid=prof_contriblnk

[日本語]
変分オートエンコーダ(Variational Auto Encoder)と手書き数字のデータセットを用いて、数字を自動生成します。
画像生成のための変数を連続的に変化させることで、
ある数字からある数字まで段階的に変化する図を得ることができます。
詳細はサムネイルをクリックしてください。GIF画像が表示されます。
例えば、異常検知のためにVAEを利用可能です。
https://jp.mathworks.com/matlabcentral/fileexchange/73283-anomaly-detection-using-variational-autoencoder-vae?s_tid=prof_contriblnk

[Key words]
deep learning, generative model, hand-drawn, hand-written, MINIST, reconstruction, VAE, variational auto encoder
Other reference
[1] 巣籠 悠輔, 詳説ディープラーニング: 生成モデル編 Kindle版 (amazon)
[2] Federico Errica, "Step-By-Step Derivation of SNE and t-SNE gradients"
http://pages.di.unipi.it/errica/assets/files/sne_tsne.pdf#search='StepByStep+Derivation+of+SNE+and+tSNE+gradients'
[3] Dr Stephen Odaibo: Variational Inference & Derivation of the Variational Autoencoder (VAE) Loss Function: A True Story
https://towardsdatascience.com/variational-inference-derivation-of-the-variational-autoencoder-vae-loss-function-a-true-story-3543a3dc67ee
[4] 高校数学の美しい物語:対数和不等式の証明と応用 https://mathtrain.jp/logsumineq
[5] van der Maaten, Laurens, and Geoffrey Hinton. Visualizing Data using t-SNE. J. Machine Learning Research 9, 2008, pp. 2579–2605.

Cite As

Kenta (2023). Generating digits by interpolating latent space with VAE (https://www.mathworks.com/matlabcentral/fileexchange/74450-generating-digits-by-interpolating-latent-space-with-vae), MATLAB Central File Exchange. Retrieved .

MATLAB Release Compatibility
Created with R2019b
Compatible with any release
Platform Compatibility
Windows macOS Linux

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!
Version Published Release Notes
1.4

Keywords added

1.3

Explanation added

1.2

Description added

1.1

efficiency improved

1.0.1

Title changed

1.0.0