Implementation of ToeplitzLDA for spatiotemporal stationary time series data.

Overview

ToeplitzLDA

Code for the ToeplitzLDA classifier proposed in here. The classifier conforms sklearn and can be used as a drop-in replacement for other LDA classifiers. For in-depth usage refer to the learning from label proportions (LLP) example or the example script.

Note we used Ubuntu 20.04 with python 3.8.10 to generate our results.

Getting Started / User Setup

If you only want to use this library, you can use the following setup. Note that this setup is based on a fresh Ubuntu 20.04 installation.

Getting fresh ubuntu ready

apt install python3-pip python3-venv

Python package installation

In this setup, we assume you want to run the examples that actually make use of real EEG data or the actual unsupervised speller replay. If you only want to employ ToeplitzLDA in your own spatiotemporal data / without mne and moabb then you can remove the package extra neuro, i.e. pip install toeplitzlda or pip install toeplitzlda[solver]

  1. (Optional) Install fortran Compiler. On ubuntu: apt install gfortran
  2. Create virtual environment: python3 -m venv toeplitzlda_venv
  3. Activate virtual environment: source toeplitzlda_venv/bin/activate
  4. Install toeplitzlda: pip install toeplitzlda[neuro,solver], if you dont have a fortran compiler: pip install toeplitzlda[neuro]

Check if everything works

Either clone this repo or just download the scripts/example_toeplitz_lda_bci_data.py file and run it: python example_toeplitz_lda_bci_data.py. Note that this will automatically download EEG data with a size of around 650MB.

Alternatively, you can use the scripts/example_toeplitz_lda_generated_data.py where artificial data is generated. Note however, that only stationary background noise is modeled and no interfering artifacts as is the case in, e.g., real EEG data. As a result, the overfitting effect of traditional slda on these artifacts is reduced.

Using ToeplitzLDA in place of traditional shrinkage LDA from sklearn

If you have already your own pipeline, you can simply add toeplitzlda as a dependency in your project and then replace sklearns LDA, i.e., instead of:

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
clf = LinearDiscriminantAnalysis(solver="lsqr", shrinkage="auto")  # or eigen solver

use

from toeplitzlda.classification import ToeplitzLDA
clf = ToeplitzLDA(n_channels=your_n_channels)

where your_n_channels is the number of channels of your signal and needs to be provided for this method to work.

If you prefer using sklearn, you can only replace the covariance estimation part, note however, that this in practice (on our data) yields worse performance, as sklearn estimates the class-wise covariance matrices and averages them afterwards, whereas we remove the class-wise means and the estimate one covariance matrix from the pooled data.

So instead of:

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
clf = LinearDiscriminantAnalysis(solver="lsqr", shrinkage="auto")  # or eigen solver

you would use

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from toeplitzlda.classification.covariance import ToepTapLW
toep_cov = ToepTapLW(n_channels=your_n_channels)
clf = LinearDiscriminantAnalysis(solver="lsqr", covariance_estimator=toep_cov)  # or eigen solver

Development Setup

We use a fortran compiler to provide speedups for solving block-Toeplitz linear equation systems. If you are on ubuntu you can install gfortran.

We use poetry for dependency management. If you have it installed you can simply use poetry install to set up the virtual environment with all dependencies. All extra features can be installed with poetry install -E solver,neuro.

If setup does not work for you, please open an issue. We cannot guarantee support for many different platforms, but could provide a singularity image.

Learning from label proportions

Use the run_llp.py script to apply ToeplitzLDA in the LLP scenario and create the results file for the different preprocessing parameters. These can then be visualized using visualize_llp.py to create the plots shown in our publication. Note that running LLP takes a while and the two datasets will be downloaded automatically and are approximately 16GB in size. Alternatively, you can use the results provided by us that are stored in scripts/usup_replay/provided_results by moving/copying them to the location that visualize_llp.py looks for.

ERP benchmark

This is not yet available.

Note this benchmark will take quite a long time if you do not have access to a computing cluster. The public datasets (including the LLP datasets) total a size of approximately 120GB.

BLOCKING TODO: How should we handle the private datasets?

FAQ

Why is my classification performance for my stationary spatiotemporal data really bad?

Check if your data is in channel-prime order, i.e., in the flattened feature vector, you first enumerate over all channels (or some other spatially distributed sensors) for the first time point and then for the second time point and so on. If this is not the case, tell the classifier: e.g. ToeplitzLDA(n_channels=16, data_is_channel_prime=False)

Owner
Jan Sosulski
Jan Sosulski
High-Resolution Image Synthesis with Latent Diffusion Models

Latent Diffusion Models arXiv | BibTeX High-Resolution Image Synthesis with Latent Diffusion Models Robin Rombach*, Andreas Blattmann*, Dominik Lorenz

CompVis Heidelberg 5.6k Dec 30, 2022
Hub is a dataset format with a simple API for creating, storing, and collaborating on AI datasets of any size.

Hub is a dataset format with a simple API for creating, storing, and collaborating on AI datasets of any size. The hub data layout enables rapid transformations and streaming of data while training m

Activeloop 5.1k Jan 08, 2023
Minimal deep learning library written from scratch in Python, using NumPy/CuPy.

SmallPebble Project status: experimental, unstable. SmallPebble is a minimal/toy automatic differentiation/deep learning library written from scratch

Sidney Radcliffe 92 Dec 30, 2022
Repository for the COLING 2020 paper "Explainable Automated Fact-Checking: A Survey."

Explainable Fact Checking: A Survey This repository and the accompanying webpage contain resources for the paper "Explainable Fact Checking: A Survey"

Neema Kotonya 42 Nov 17, 2022
Trained on Simulated Data, Tested in the Real World

Trained on Simulated Data, Tested in the Real World

livox 43 Nov 18, 2022
Code and real data for the paper "Counterfactual Temporal Point Processes", available at arXiv.

counterfactual-tpp This is a repository containing code and real data for the paper Counterfactual Temporal Point Processes. Pre-requisites This code

Networks Learning 11 Dec 09, 2022
Official Repository for "Robust On-Policy Data Collection for Data Efficient Policy Evaluation" (NeurIPS 2021 Workshop on OfflineRL).

Robust On-Policy Data Collection for Data-Efficient Policy Evaluation Source code of Robust On-Policy Data Collection for Data-Efficient Policy Evalua

Autonomous Agents Research Group (University of Edinburgh) 2 Oct 09, 2022
pytorch implementation of Attention is all you need

A Pytorch Implementation of the Transformer: Attention Is All You Need Our implementation is largely based on Tensorflow implementation Requirements N

230 Dec 07, 2022
Pytorch implementations of the paper Value Functions Factorization with Latent State Information Sharing in Decentralized Multi-Agent Policy Gradients

LSF-SAC Pytorch implementations of the paper Value Functions Factorization with Latent State Information Sharing in Decentralized Multi-Agent Policy G

Hanhan 2 Aug 14, 2022
pytorch bert intent classification and slot filling

pytorch_bert_intent_classification_and_slot_filling 基于pytorch的中文意图识别和槽位填充 说明 基本思路就是:分类+序列标注(命名实体识别)同时训练。 使用的预训练模型:hugging face上的chinese-bert-wwm-ext 依

西西嘛呦 33 Dec 15, 2022
Source code for the NeurIPS 2021 paper "On the Second-order Convergence Properties of Random Search Methods"

Second-order Convergence Properties of Random Search Methods This repository the paper "On the Second-order Convergence Properties of Random Search Me

Adamos Solomou 0 Nov 13, 2021
[TIP2020] Adaptive Graph Representation Learning for Video Person Re-identification

Introduction This is the PyTorch implementation for Adaptive Graph Representation Learning for Video Person Re-identification. Get started git clone h

WuYiming 41 Dec 12, 2022
Creating Artificial Life with Reinforcement Learning

Although Evolutionary Algorithms have shown to result in interesting behavior, they focus on learning across generations whereas behavior could also be learned during ones lifetime.

Maarten Grootendorst 49 Dec 21, 2022
Original code for "Zero-Shot Domain Adaptation with a Physics Prior"

Zero-Shot Domain Adaptation with a Physics Prior [arXiv] [sup. material] - ICCV 2021 Oral paper, by Attila Lengyel, Sourav Garg, Michael Milford and J

Attila Lengyel 40 Dec 21, 2022
CryptoFrog - My First Strategy for freqtrade

cryptofrog-strategies CryptoFrog - My First Strategy for freqtrade NB: (2021-04-20) You'll need the latest freqtrade develop branch otherwise you migh

Robert Davey 137 Jan 01, 2023
Playable Video Generation

Playable Video Generation Playable Video Generation Willi Menapace, Stéphane Lathuilière, Sergey Tulyakov, Aliaksandr Siarohin, Elisa Ricci Paper: ArX

Willi Menapace 136 Dec 31, 2022
Symbolic Music Generation with Diffusion Models

Symbolic Music Generation with Diffusion Models Supplementary code release for our work Symbolic Music Generation with Diffusion Models. Installation

Magenta 119 Jan 07, 2023
Causal Influence Detection for Improving Efficiency in Reinforcement Learning

Causal Influence Detection for Improving Efficiency in Reinforcement Learning This repository contains the code release for the paper "Causal Influenc

Autonomous Learning Group 21 Nov 29, 2022
Apollo optimizer in tensorflow

Apollo Optimizer in Tensorflow 2.x Notes: Warmup is important with Apollo optimizer, so be sure to pass in a learning rate schedule vs. a constant lea

Evan Walters 1 Nov 09, 2021
SMPL-X: A new joint 3D model of the human body, face and hands together

SMPL-X: A new joint 3D model of the human body, face and hands together [Paper Page] [Paper] [Supp. Mat.] Table of Contents License Description News I

Vassilis Choutas 1k Jan 09, 2023