t-SNE Dimensionality Reduction with Scikit-Learn

[@annyd3406]

no one explains t-sne thank you that's a rare video !!!!

Thanks for your comment, Anny!
I'm glad you found this explanation useful. =)
Kindly,
Nono

[@jackiexu3683]

That is amazing! May I know do we have the link for the jupyter notebook or google colab file?

Hi, Jackie!
🐍 Colab Notebook › colab.research.google.com/drive/1vKFJehl-o2AhOajz9fxxwiIbQ4jwp6Vc?usp=sharing
Thanks for the comment, I forgot to link the notebook from the video notes.
Let me know if that works!
Kindly,
Nono

[@jackiexu3683]

Detailed Summary:
[00:00](ruclips.net/video/DtFQAJmlID0/видео.html) Tutorial on dimensionality reduction using t-SNE with scikit-learn
- Using stacked autoencoders to obtain codings for input images
- Visualizing the reduced dimensions with t-SNE and plotting for understanding similarities
[09:23](ruclips.net/video/DtFQAJmlID0/видео.html) Loading and processing small image dataset in TensorFlow.
- Data gets deleted when the runtime is recycled on Colab.
- Process small image dataset using TensorFlow and load in the same format as larger dataset.
[17:00](ruclips.net/video/DtFQAJmlID0/видео.html) Loading and preprocessing dataset for dimensionality reduction
- Batching and converting RGB images to grayscale
- Loading manually using Glob and PIL libraries
[24:08](ruclips.net/video/DtFQAJmlID0/видео.html) Implemented encoder-decoder architecture for image classification
- Trained the model with 40 epochs and achieved 98.16% accuracy
- Generated images using the trained model and plotted input and output images
[31:09](ruclips.net/video/DtFQAJmlID0/видео.html) Implemented and trained a deep learning model for image generation
- The model was trained on a custom dataset and achieved 99.32% accuracy
- Generated images using the trained decoder and observed changes in codings
[37:29](ruclips.net/video/DtFQAJmlID0/видео.html) Generative model trained for longer refines results
- Training for more epochs shows improvements in codings
- Small changes in codings significantly alter resulting image
[44:08](ruclips.net/video/DtFQAJmlID0/видео.html) Testing clustering and dimensionality reduction algorithms
- Experimented with t-SNE and plotted results in scatterplot
- Visualized scatterplot with images to see clustering of drawings
[51:23](ruclips.net/video/DtFQAJmlID0/видео.html) Visualizing similarity on Fashion MNIST
- t-SNE plots can group similar images together
- This can be used for recommendation systems

That's awesome - did you auto-generate it?

[@linda_erose]

is there a more basic version of tsne? this seems like a lot for me as a beginner

I don't think I have other videos that make this easier. Let us know if you find any sources though! Here's a quick solution from ChatGPT 4o that may work on a single Python file. =)
chatgpt.com/share/22f3b948-2a7f-4555-8326-1cb4de22e7a2