Speaker Detail

Visualizing High-Dimensional Data

Data usually have high dimensionality, especially unstructured data. Extracting [unigrams](https://en.wikipedia.org/wiki/N-gram) from text data, for example, usually yields hundreds to thousands of features. Even when we are using machine learning models to do a supervised learning task, interpretability of the model is usually hard to achieve.

A workaround to this is to visualize the data in 2D, so that we can see some interesting patterns in the data. I will show some interesting figures from applying dimensionality reduction techniques, such as [PCA](https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html), [autoencoder](https://blog.keras.io/building-autoencoders-in-keras.html), [t-SNE](https://distill.pub/2016/misread-tsne/), and [UMAP](https://github.com/lmcinnes/umap) to some data. I will also introduce [Altair](https://altair-viz.github.io/) as the go to library to interact with the resulting visualization.