Quick testing of the Jensen-Shannon implementation in the philentropy package

100%

Quick testing of the Jensen-Shannon implementation in the philentropy package

✓ Published0🌍 Public

Last edited Nov 9, 2018

Created on Nov 9, 2018

This example tests the Jensen-Shannon divergence (JSD) and distance calculations in the `philentropy` R package, using toy distributions to validate the expected behavior for positive and negative controls. It shows that perfectly overlapping distributions yield a distance of 1, while fully independent distributions yield a distance of 0. The code manually normalizes two input vectors (regulon activity scores and cell identity indicators) and passes them as a matrix to `philentropy::JSD`, then applies the formula `1 - sqrt(JSD)` to derive the distance. The source is a public gist by FloWuenne, demonstrating a quick sanity check for the package's implementation.

AI-generated description

Similar vizzes

Loading thumbnail…

Normal Distributions

This example visualizes two normal distributions as filled, semi-transparent density curves overlaid on the same axes. It shows how the spread and peak of the distributions differ based on their standard deviations. The code uses D3 v4 with SVG rendering, and computes the probability densities using the `jStat` library’s `normal.pdf` function. Random data is generated by the custom `Random_normal_Dist` function, which samples quantiles from a mean and standard deviation. The `d3.line` generator draws the curves, and the horizontal axis is scaled with `d3.scaleLinear`.

EEfratVil

49% match

Loading thumbnail…

Untitled

This example visualizes the distribution of probability estimates from eight verbal uncertainty phrases (such as “Almost Certainly” and “Almost No Chance”) using kernel density estimation. Each smoothed density curve is colored with a viridis scale based on the group mean, revealing differences in central tendency and spread across categories. The visualization uses D3 v4 to load CSV data from GitHub, compute the means, and render the densities as SVG paths. The kernel density estimator applies an Epanechnikov kernel, while axes, gridlines, and labels are drawn with `d3.axisBottom`, `d3.axisLeft`, and the `d3.scaleSequential` with `d3.interpolateViridis`.

SSimone

45% match

Loading thumbnail…

Quick testing of CIDR implementation

This example tests a CIDR implementation for single-cell RNA sequencing data, specifically using colorectal cancer cells from a 2017 Li et al. study. It first reads in the data and applies the CIDR clustering workflow, with most steps commented out. The visualization shows a two-dimensional principal component analysis (PCA) plot, coloring points by cell type and using plotting symbols to denote CIDR clusters. The code uses the `scDataConstructor`, `determineDropoutCandidates`, and `plot` functions from the CIDR package, alongside `ggplot2` for an alternative scatterplot of the same PC axes.

FFloWuenne

44% match

Loading thumbnail…

Visualizing Distributions

This example compares four chart types for visualizing one-dimensional distributions. It first displays a swarm plot and a pile plot, which use a force-directed layout to position individual data points without overlap, in this case 700 values generated from a log-normal distribution. The example then shows a histogram, constructed using the `d3.layout.histogram()` function to bin the data, and a boxplot, which uses `d3.scale.quantile()` to compute summary statistics. The swarm and pile plots are built using the `d3.forceChart()` plugin from the `force-chart.js` library.

BBBischof

43% match

Loading thumbnail…

Testing jStat

This visualization displays average hemoglobin levels by sex, with each horizontal bar showing the mean and a 95% confidence interval computed using the `jStat.tci` function. The chart uses React and D3 v6, with `scaleBand` for the categorical y-axis and `scaleLinear` for the numeric x-axis, rendered as SVG through a reusable `VizContainer`. Data is loaded from a CSV file via `d3.csv`, and the bars include optional confidence-interval lines that are drawn from the lower to upper bound of the interval.

WWaseem Medhat

43% match

Loading thumbnail…

Mahalanobis Distance

A slider adjusts the correlation level between two variables, and the scatter plot updates to show how the Mahalanobis distance of each point changes accordingly. The visualization encodes distance through color, transitioning from green for low distances to red for high ones, overlaid on 2,000 points sampled from a bivariate normal distribution. The code uses D3 v4 to render an SVG with axes and circles, while the Mahalanobis distance is computed via a custom Cholesky decomposition and matrix operations, with colors interpolated using a linear scale with HCL color space.

EEfratVil

43% match

Loading thumbnail…

Untitled

This example shows the distribution of two variables as overlapping histograms, with semi-transparent bars in teal and dark purple to reveal how their ranges compare. The visualization is built with D3.js v4 and rendered as SVG, using `d3.histogram()` to bin the data and `d3.scaleLinear()` for the axes. The code loads a CSV file, filters it by a `type` column, and draws both series on a shared x-axis while scaling the y-axis to the maximum bin count. A handmade legend identifies the two variables.

JJens Grochtdreis

43% match

Loading thumbnail…

Contour Test using Jason Davis implementation with D3

This visualization demonstrates contour generation on a 20×10 grid of numeric values using the Conrec algorithm, ported to D3. The scalar field is encoded as a series of filled contour polygons, each mapped to a distinct level between -5 and 3. A linear color scale interpolates from white to red, with semi-transparent fills and black strokes delineating the contours. The data is rendered into an SVG using D3's line generator, with spatial scaling applied to map the grid coordinates onto a 500×500 pixel canvas. The example showcases how D3 can work alongside external contouring libraries to produce smooth isolines directly in the browser.

DDinoJay

42% match