Han Liu @ Computer Science, Northwestern University

Current Highlight

The MAGICS Lab Github

Maintainers: current members of the MAGICS lab at Northwestern University.

image

Repository Description: The MAGICS Lab Github hosts open sourced models and code developed and maintained by the current MAGICS lab members. It includes the source code for DNABERT, DNABERT-2, and DNABERT-S, SparseModernHopfield, and STanHop.

» Learn more

Highlights

DNABERT, DNABERT-2, DNABERT-S.

Important Features

  • 1 Large language model for genome.
  • 2 More than 2 million times downloads.
  • 3 500+ stars for DNABERT and 150+ for DNABERT-2.
  • 4 Species-aware DNA embedding.
  • 5 Light-weight and high performance.

  • HUGE: High Dimensional Undirected Graphical Model Estimation

    Maintainers: Tuo Zhao and Han Liu.

    image

    Software Description: The package "huge" provides a general framework for high-dimensional undirected graph estimation. It integrates data preprocessing, neighborhood screening, graph estimation, and model selection techniques into a pipeline.

    » Learn more

    Highlights

    Estimating nonparanormal graphical models.

    Important Features

  • 1 Instead of using Fortan, it is written in C.
  • 2 Fitting high dimensional Gausian copula models.
  • 3 Graph generation, data generation and visualization.
  • 4 Support lossless and lossy screening rules.
  • 5 Subsampling and Permutation based model selection.
  • 6 Sparse matrix representation for scalability.
  • fastclime: Parametric Simplex Algorithm for Relaxed LP Problems

    Maintainers: Haotian Pang and Han Liu.

    image

    Software Description: The package "fastclime" provides a method of recover the precision matrix efficiently by applying parametric simplex method. The computation is based on linear optimization. It also contains a generic LP solver and a parameterized LP solver using parametric simplex method.

    » Learn more

    Highlights

    Efficient implementation of parametric simplex algorithm.

    Important Features

  • 1 Completely written in C with a R wrapper.
  • 2 Pathwise estimation of sparse precision matrix.
  • 3 Generic parametric simplex solver for relaxed LP.
  • 4 Exact dual certificate as stopping criterion.
  • 5 Efficient pathwise optimization.
  • 6 Implements the CLIME algorithm for graph estimation.
  • SMART

    Sparse Multivariate Analysis via Rank Transformation.

    image

    » Download

    Bigmatrix

    Tuning Insensitive Graph Estimation and Regression.

    image

    » Download

    SAM

    Sparse Additive Machine.

    image

    » Download

    BigTSP

    Top Scoring Pair based methods for classification.

    image

    » Download

    Software to come

    Function 2.

    image

    » Download

    Software to come

    Function 3.

    image

    » Download

    Reading Group

    We have weekly reading group. The topics of this quarter include large language models, multivariate time series forecasting, generative AI.

    Get In Touch

    Department of Computer Science
    Mudd Hall 3119
    Northwestern University
    Evanston, IL 60201
    Phone: +847 491 2793
    Email: hanliu@northwestern.edu