Call event detections using activation thresholds.

Converts raw neuron activation values into binary detections by applying a numeric threshold to each named neuron. Frames where the activation exceeds the threshold are marked TRUE; all others are FALSE. All activation_ columns are dropped from the output — use bin afterwards to summarise detection counts over time.

Usage

call_detections(results, thresholds)

Arguments

results

A data frame or data.table of buzzdetect results, such as created by read_results.. Must contain activation_ columns for each neuron named in thresholds.

thresholds

A named numeric vector mapping neuron names to detection thresholds (e.g. c(ins_buzz = -1.2)). Frames whose activation value exceeds the threshold are counted as detections. Because model_general_v3 outputs negative log-likelihoods, thresholds are typically negative. Activations above the threshold value are counted as detections, so thresholds for models that output negative log-likelihoods (such as model_general_v3) are typically negative (e.g. -1.2).

Value

A data.table with detections_ columns replacing the activation_ columns. Each detection column contains logical (TRUE/FALSE) values.

Details

The original data frame is never modified; a copy is returned.

See also

bin to count detections per time bin, bin_directory to run the full pipeline in one call.

Examples

path <- system.file(
  'extdata/five_flowers/soybean/9/230809_0000_buzzdetect.csv',
  package = 'buzzr'
)
results <- read_results(path)

# Single neuron: activations above -1.2 are counted as buzz detections
call_detections(results, thresholds = c(ins_buzz = -1.2))
#>        start_filetime detections_ins_buzz
#>                 <num>              <lgcl>
#>     1:           0.00               FALSE
#>     2:           0.96               FALSE
#>     3:           1.92               FALSE
#>     4:           2.88               FALSE
#>     5:           3.84               FALSE
#>    ---                                   
#> 89996:       86395.20               FALSE
#> 89997:       86396.16               FALSE
#> 89998:       86397.12               FALSE
#> 89999:       86398.08               FALSE
#> 90000:       86399.04               FALSE

# Multiple neurons at once
call_detections(results, thresholds = c(ins_buzz = -1.2, mech_plane = -2.0))
#>        start_filetime detections_ins_buzz detections_mech_plane
#>                 <num>              <lgcl>                <lgcl>
#>     1:           0.00               FALSE                  TRUE
#>     2:           0.96               FALSE                 FALSE
#>     3:           1.92               FALSE                 FALSE
#>     4:           2.88               FALSE                 FALSE
#>     5:           3.84               FALSE                 FALSE
#>    ---                                                         
#> 89996:       86395.20               FALSE                 FALSE
#> 89997:       86396.16               FALSE                 FALSE
#> 89998:       86397.12               FALSE                 FALSE
#> 89999:       86398.08               FALSE                 FALSE
#> 90000:       86399.04               FALSE                 FALSE