epiout.negative_binomial

Module Contents

Classes

NegativeBinomial

Negative binomial distribution with mean and dispersion parameters.

NBLoss

Negative binomial loss function.
class epiout.negative_binomial.NegativeBinomial(mean=None, dispersion=None, dispersion_min=0.1, dispersion_max=1000.0, mean_min=0.5)

Negative binomial distribution with mean and dispersion parameters.

Parameters

  • mean – Mean parameter of negative binomial distribution.

  • dispersion – Dispersion parameter of negative binomial distribution.

  • dispersion_min – Minimum possible value of dispersion parameter.

  • dispersion_max – Maximum possible value of dispersion parameter.

  • mean_min – Minimum possible value of mean parameter.

infer_dispersion(self, counts, method='mom', mle_kwargs=None)

Infer dispersion parameter from counts.

Parameters

  • counts – Counts of peaks.

  • method – Method to infer dispersion parameter. Valid values are mom and mle.

infer_mean(self, counts, method='mom')

Infer mean parameter from counts.

_infer_mean_mom(self, counts)
_infer_dispersion_mom(self, counts)
_infer_dispersion_mle(self, counts, iteration=10, decay_rate=0.5, c1=10 ** - 4, c2=0.9)
pval(self, counts)
Calculate p-value from counts and inferred mean

and dispersion parameters.

log_prob(self, counts, eps=1e-08)

Log probability of counts given mean and dispersion parameters.

static _log_prob(counts, mean, dispersion, eps=1e-08)
class epiout.negative_binomial.NBLoss(dispersion_min=0.1, dispersion_max=1000.0)

Bases: tensorflow.keras.losses.Loss

Negative binomial loss function.

Parameters

  • dispersion_min – Minimum possible value of dispersion parameter.

  • dispersion_max – Maximum possible value of dispersion parameter.

call(self, y_true, y_pred)

Invokes the Loss instance.

Parameters

  • y_true – Ground truth values. shape = [batch_size, d0, .. dN], except sparse loss functions such as sparse categorical crossentropy where shape = [batch_size, d0, .. dN-1]

  • y_pred – The predicted values. shape = [batch_size, d0, .. dN]

Returns

Loss values with the shape [batch_size, d0, .. dN-1].