Supervised learning function for temporal decoding.
Project description
Supervised Learning Toolbox for Neural data
This toolbox facilitates neural decoding of time series. Under the hood it uses scikit-learn functions.
Set-up
Install using pip
pip install temp_dec
Requirements
sklearn==0.21.3
numpy==1.13.1
scipy==0.19.1
Example
This function takes in the data X (ndarray; trials by channels by time), labels y (ndarray; vector), and a time (ndarray, vector).
from temp_dec import decoding_functions
Using a sliding time window
If there is information in the temporal dynamics of the signal, using a sliding time window will increase decoding accuracy (and smooth the signal). We also demean the signal within each window, this avoids the issue of baselining.
temporal_dymanics == True
Applying PCA
If you use a large amount of features, you might want to consider applying PCA to your features before applying your classifier. In addition, classifiers are sensitive to noise rejecting noise components from the data can be beneficial.
use_pca == True
You can also regulate how many components you would like to keep (setting the pca_components variant to > 1) or how much variance you would like to explain (setting the pca_components variant to < 1). As a general rule of thumb maintaining 95% of variance will maintain enough signal and reduces feature space.
pca_components == .95
Classifiers
Different classifiers are supported, selected in accordance with Grootwagers et al (2017) j.cogn.neurosci.
- LDA: linear disciminant analysis
- LG: logistic regression
- GNB: Gaussian Naive Bayes
- maha: Nearest Neighbours using mahalanobis distance.
classifier == 'LDA'
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for temp_dec-1.0.5-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 646b747638556fd9d89e6ab67e49f2d312c7719cb3f2fe3116a1ef9e7bc89cdf |
|
MD5 | dc24c0325f2bd9334e6a14c2ef6b50fe |
|
BLAKE2b-256 | 1ceadf962cdf140f2fa8a4cd109d5a3faf3c65cc034450fde159e750244170cf |