View source on GitHub |
Statistical functions.
Functions
assign_log_moving_mean_exp(...)
: Compute the log of the exponentially weighted moving mean of the exp.
assign_moving_mean_variance(...)
: Compute one update to the exponentially weighted moving mean and variance.
auto_correlation(...)
: Auto correlation along one axis.
brier_decomposition(...)
: Decompose the Brier score into uncertainty, resolution, and reliability.
brier_score(...)
: Compute Brier score for a probabilistic prediction.
cholesky_covariance(...)
: Cholesky factor of the covariance matrix of vector-variate random samples.
correlation(...)
: Sample correlation (Pearson) between observations indexed by event_axis
.
count_integers(...)
: Counts the number of occurrences of each value in an integer array arr
.
covariance(...)
: Sample covariance between observations indexed by event_axis
.
cumulative_variance(...)
: Cumulative estimates of variance.
expected_calibration_error(...)
: Compute the Expected Calibration Error (ECE).
expected_calibration_error_quantiles(...)
: Expected calibration error via quantiles(exp(pred_log_prob),num_buckets)
.
find_bins(...)
: Bin values into discrete intervals.
histogram(...)
: Count how often x
falls in intervals defined by edges
.
kendalls_tau(...)
: Computes Kendall's Tau for two ordered lists.
log_average_probs(...)
: Computes log(average(to_probs(logits)))
in a numerically stable manner.
log_loomean_exp(...)
: Computes the log-leave-one-out-mean of exp(logx)
.
log_loosum_exp(...)
: Computes the log-leave-one-out-sum of exp(logx)
.
log_soomean_exp(...)
: Computes the log-swap-one-out-mean of exp(logx)
.
log_soosum_exp(...)
: Computes the log-swap-one-out-sum of exp(logx)
.
moving_mean_variance_zero_debiased(...)
: Compute zero debiased versions of moving_mean
and moving_variance
.
percentile(...)
: Compute the q
-th percentile(s) of x
.
quantiles(...)
: Compute quantiles of x
along axis
.
stddev(...)
: Estimate standard deviation using samples.
variance(...)
: Estimate variance using samples.
windowed_mean(...)
: Windowed estimates of mean.
windowed_variance(...)
: Windowed estimates of variance.