How to use HyperTools

Plot

• Plotting
  • Import Hypertools and other libraries for tutorial
  • Load your data
  • Plot with default settings
  • Plot in 2D
  • Using other dimensionality reduction algorithms
  • Coloring by hue
  • Interpolating missing data
  • Labeling plotted points
  • Data Clustering
  • Normalization
  • Aligning datasets with different coordinate systems
  • Saving
  • Plotting text using semantic models

Analyze

• Analyze
  • Import packages
  • Load your data
  • Normalization
  • Normalize and reduce
  • Normalize, reduce, and align

Normalize

• Normalization
  • Import packages
  • Generate synthetic data
  • Visualize the data
  • Normalizing (Specified Cols or Rows)

Reduce

• Dimensionality reduction
  • Import Hypertools
  • Load your data
  • Reduce one array
  • Reduce list of arrays
  • Reduce list of arrays (TSNE)
  • Reduce to specified number of dimensions
  • Reduce list of arrays with specific parameters

Align

• Alignment
  • Import Hypertools
  • Load your data
  • Visualize unaligned data
  • Aligning data with Hyperalignment
  • Aligning data with the Shared Response Model

Cluster

• Clustering with Hypertools
  • Import Packages
  • Load your data
  • Obtain cluster labels
  • Specify number of cluster labels
  • Different clustering models

Plotting text

• Visualizing text
  • Visualizing NIPS papers
  • Visualizing Wikipedia pages
  • Visualizing State of the Union addresses
  • Changing the reduction model
  • Defining a corpus

Visualizing Hugging Face embeddings

• Visualizing Hugging Face text embeddings
  • Embedding news headlines
  • Plotting the embedding space
  • Soft clustering with Gaussian mixture models
  • Nonlinear embeddings with UMAP
  • Animated 3D visualization
  • Next steps

Modern scikit-learn models and dynamics

• Modern scikit-learn models and dynamical systems
  • Density-based clustering with HDBSCAN
  • Soft clustering with Gaussian mixture models
  • Visualizing dynamics: the Lorenz attractor
  • Animating a trajectory
  • Next steps

Mapping Wikipedia with modern text embeddings

• Mapping Wikipedia with modern text embeddings
  • Loading the sample Wikipedia corpus
  • Embedding the articles
  • Mapping the corpus with UMAP and soft clustering
  • Taking a spin around the map
  • Live keywords: mapping fresh articles by category (GH #187)
  • Next steps

Visualizing the shape of a conversation

• Visualizing the shape of a conversation
  • Downloading a corpus of Reddit conversations
  • Choosing a conversation
  • Embedding each utterance with a sliding sentence window
  • One color per speaker
  • The static shape of the conversation
  • Watching the conversation unfold
  • Next steps

Plotting streaming data

• Plotting streaming data
  • Streaming from a generator
  • Comet-style display with stream_window
  • Streaming a Hugging Face dataset

Streaming from a Lab Streaming Layer (LSL) device

• Streaming from a Lab Streaming Layer (LSL) device
  • A synthetic LSL outlet
  • Resolving the stream and plotting it live
  • Cleaning up the synthetic outlet
  • Using a real device

Forecasting stock prices with hyp.predict

• Forecasting stock prices with hyp.predict
  • Downloading real price data
  • Honest backtesting: hold out the last 30 trading days
  • Visualizing forecasts in reduced space
  • Best model vs. actual, in price space
  • Reusing a fitted forecaster with return_model=True
  • Next steps

Imputing and forecasting a real projectile arc with hyp.impute and hyp.predict

• Imputing and forecasting a real projectile arc with hyp.impute and hyp.predict
  • Why a Kalman filter for projectile motion?
  • Downloading and extracting the real tracking data
  • Part 1: Imputation – filling occluded and scattered gaps
  • Part 2: Forecasting – extrapolating the arc’s remainder
  • Honest takeaways
  • Next steps

Story trajectories: brain activity while listening to a story

An animated cloud of hyperaligned trajectories showing how all 36 subjects’ whole-brain activity traces out a shared path through a low-dimensional space while they listen to the same spoken story (fMRI data from Simony et al., 2016). Each subject is preprocessed with a per-subject manip (Smooth → Resample → ZScore), hyperaligned in the 100-hub feature space (n_iter=10) and only then reduced to 3-D with reduce='IncrementalPCA'; an animate='window' trail then slides along the aligned trajectories so you watch all 36 subjects move together through the story. Aligning in the hub space before reducing (rather than over-reducing first, which starves hyperalignment) is what pulls the trajectories together – their within-timepoint spread tightens ~30%. See the full gallery example, Story trajectories: brain activity while listening to a story.