csvdimreduce

ARGS:
    <columns>
      List of columns to use as coordinates. First column is number 1. Parsing support ranges with steps like 3,4,10:5:100.
      See `number_range` Rust crate for details.
      Use `xsv headers your_file.csv` to find out column numbers.

    <n_out_coords>
      Number of output coordinates (new fields in CSV containing computed values)

      This includes temporary coordinates used for squeezing (-S).

    [path]
      Input csv file. Use stdin if absent.

OPTIONS:
    --save-each-n-iters <n>

    --no-header
      First line of the CSV is not headers

    --no-output-header
      Do not output CSV header even though input has headers

    --delimiter <delimiter>
      Field delimiter in CSV files. Comma by default.

    --record-delimiter <delimiter>
      Override line delimiter in CSV files.

    -o, --output <path>
      Save file there instead of stdout

    --random-seed <seed>
      Initial particle positions

    -w, --weight <column_number>
      Use this column as weights

    -n, --n-iters <n>
      Basic number of iterations. Default is 100.
      Note that complexity of each iteration is quadratic of number of lines in CSV.

    -r, --rate <rate>
      Initial rate of change i.e. distance the fastest particle travels per iteration.
      Default is 0.01.

    --inertia-multiplier <x>
      Apply each movement multiplpe times, decaying it by this factor. Default is 0.9.

    -R, --final-rate <final_decay>
      Ramp down rate of change to this value at the end.

    -c, --central-force <f>
      Attract particles' coordinates to 0.5 with this strenght (relative to average inter-particle forces).

    -F, --same-particle-force <f>
      Additional repelling force between particles (even those with the same parameters). Default is 0.2

    -S, --retain_coords_from_squeezing <n>
      After doing usual iterations, perform additional steps to \"flatten\" the shape into fewer dimension count (squeeze phase).
      Specified number of coodinaes are retained. For others, the `-c` central force is crancked up to `-C`, so they
      (should) become flat \"0.5\" in the end.
      This produces better results compared to just having that number of coordinates from the beginning.

    --squeeze-rampup-rate <rate>
      Use this `-r` rate when doing the squeeze phase.

    --squeeze-rampup-iters <n>
      User this number of iterations of the first phase of squeezing phase.
      This applies to each squeezed dimension sequentially.

    -C, --squeeze-final-force <f>
      This this central force for the squeezed dimensions.
      The force is gradually increased from `-C` to this value during the rampup phase.

    --squeeze-final-initial-rate <rate>
      Override `-r` rate for the second phase of squeezing. It decays with `-d` each iteration.

    --squeeze-final-iters <n>
      Number of iterations of the second phase of squeezeing (where central force no longer changes, just to increase precision)

    --warmup-iterations <n>
      Gradually increase rate from zero during this number of iterations. Defaults to 10.

    --debug
      Print various values, including algorithm parameter values

    -N, --normalize
      Automatically normalize the data

    -h, --help
      Prints help information.