Implementation of the ivis algorithm as described in the paper Structure-preserving visualisation of high dimensional single-cell datasets. Ivis is designed to reduce dimensionality of very large datasets using a siamese neural network trained on triplets. Both unsupervised and supervised modes are supported.
Install the latest ivis release from PyPi:
pip install ivis
Alternatively, you can install the development version from github:
git clone https://github.com/beringresearch/ivis
cd ivis
pip install -r requirements.txt --editable .
The following optional dependencies are needed if using the visualization callbacks while training the Ivis model:
- matplotlib
- seaborn
Ivis Python package is updated frequently! To upgrade, run:
pip install ivis --upgrade
- Scalable: ivis is fast and easily extends to millions of observations and thousands of features.
- Versatile: numpy arrays, sparse matrices, and hdf5 files are supported out of the box, making it easy to apply ivis to heterogeneous problems including clustering and anomaly detection.
- Accurate: ivis excels at preserving both local and global features of a dataset. Often, ivis performs better at preserving global structure of the data than t-SNE, making it easy to visualise and interpret high-dimensional datasets.
- Generalisable: ivis supports addition of new data points to original embeddings via a
transformmethod, making it easy to incorporate ivis into standard sklearn Pipelines.
And many more! See ivis readme for latest additions and examples.
from ivis import Ivis
from sklearn.preprocessing import MinMaxScaler
from sklearn import datasets
iris = datasets.load_iris()
X = iris.data
X_scaled = MinMaxScaler().fit_transform(X)
model = Ivis(embedding_dims=2, k=15)
embeddings = model.fit_transform(X_scaled)
Ivis can be used in a wide variety of real-world applications. The Ivis Universe consists of packages that extend the core Ivis functionality.
- ivis-animate - visualise the Ivis learning process.
- ivis-explain - explain which features contribute the most to ivis embeddings.
Copyright 2019 Bering Limited