perf_eval: Compute performance of diffusion scores on a single case

In b2slab/diffusion: Diffusion scores on biological networks

Description

Function perf_eval directly compares a desired output with the scores from diffusion. It handles the possible shapes of the scores (vector, matrix, list of matrices) and gives the desired metrics.

Usage

perf_eval(
    prediction,
    validation,
    metric = list(auc = metric_fun(curve = "ROC"))
)

Arguments

prediction	smoothed scores; either a named numeric vector, a column-wise matrix whose rownames are nodes and colnames are different scores, or a named list of such matrices.
validation	target scores to which the smoothed scores will be compared to. Must have the same format as the input scores, although the number of rows may vary and only the matching rows will give a performance measure.
metric	named list of metrics to apply. Each metric should accept the form f(actual, predicted)

Value

A data frame containing the metrics for each comparable pair of output-validation.

Examples

# Using a matrix with four set of scores
# called Single, Row, Small_sample, Large_sample
data(graph_toy)
diff <- diffuse(
    graph = graph_toy,
    scores = graph_toy$input_mat,
    method = "raw")
df_perf <- perf_eval(
    prediction = diff,
    validation = graph_toy$input_mat)
df_perf

b2slab/diffusion documentation built on March 6, 2021, 2:38 a.m.