Edit model card

Model description

This is a model trained to classify pieces of neuron as axon, dendrite, soma, or glia, based only on their local shape and synapse features.The model is a linear discriminant classifier which was trained on compartment labels generated by Bethanny Danskin for 3 6x6x6 um boxes in the Minnie65 Phase3 dataset.

Intended uses & limitations

This model could be used to predict some compartment labels in mouse cortical connectomes, but it is unclear to what extent this model will generalize.

Training Procedure

The model was trained on local (level 2 cache) and synapse count features from 3 6x6x6 um boxes in the Minnie65 Phase3 dataset. These features were also locally aggregated in 5-hop neighborhood windows and concatenated to each level 2 node's features. The labels were generated by Bethanny Danskin and include axon, dendrite, soma, and glia compartments. The classification model was trained using a linear discriminant classifier.

Hyperparameters

Click to expand
Hyperparameter Value
memory
steps [('transformer', QuantileTransformer(output_distribution='normal')), ('lda', LinearDiscriminantAnalysis(n_components=3))]
verbose False
transformer QuantileTransformer(output_distribution='normal')
lda LinearDiscriminantAnalysis(n_components=3)
transformer__copy True
transformer__ignore_implicit_zeros False
transformer__n_quantiles 1000
transformer__output_distribution normal
transformer__random_state
transformer__subsample 10000
lda__covariance_estimator
lda__n_components 3
lda__priors
lda__shrinkage
lda__solver svd
lda__store_covariance False
lda__tol 0.0001

Model Plot

Pipeline(steps=[('transformer',QuantileTransformer(output_distribution='normal')),('lda', LinearDiscriminantAnalysis(n_components=3))])
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

Evaluation Results

Classification Report (overall)

type precision recall f1-score support
accuracy 0.944357 0.944357 0.944357 0.944357
macro avg 0.854825 0.917289 0.878753 31307
weighted avg 0.946879 0.944357 0.945155 31307

Classification Report (by class)

class precision recall f1-score support
axon 0.956309 0.964704 0.960488 16404
dendrite 0.928038 0.911341 0.919614 6948
glia 0.964442 0.935279 0.949636 7540
soma 0.570513 0.857831 0.685274 415

How to Get Started with the Model

[More Information Needed]

Model Card Authors

Ben Pedigo Bethanny Danskin

Downloads last month
0
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.