Maximum Entropy (MaxEnt) — interactive explainer

Where does this species occur? · The model estimates P(cell | environment) using MaxEnt with feature constraints.

Click cells to record species observations. In all views, darker = higher value: darker green = higher probability; darker amber/red/blue in the strip = higher elevation/temperature/moisture. The strip colors match the single-layer views exactly. Cells whose feature strip best matches the averages observed in your clicks receive the highest (darkest green) probability.

low prob

high prob

Entropy: 4.605 (100.0% of max)

This is the highest-entropy distribution that satisfies the constraints above.

View

Probability (MaxEnt) Elevation only Temperature only Moisture only

In all views, darker = higher value.

Active features

Elevation Temperature Moisture

Each checked feature becomes a constraint: the model must match its mean value across your observations.

Constraints

Each row: italic line = target from data; amber bar = that same target; colored bar = model prediction. At convergence the bars match.

Add observations to see constraints.

Predict the next word in: "the ____" · The model estimates P(word | context) using MaxEnt with feature constraints.

Click words to add them to the corpus. Each word shows feature badges: N nature word, noun, long. Hover a word to see which others share its features — those are outlined in blue and receive a probability boost from the same constraints.

Word probability

Vertical line = uniform (12.5%, no data). Green = above uniform; red = below.

Entropy: 2.079 (100.0% of max)

This is the highest-entropy distribution that satisfies the constraints above.

Active features

Nature word Noun Long word (≥6 chars)

Each checked feature becomes a constraint: the model must match how often that feature appeared in your corpus.

Constraints

Each row: italic line = target proportion from corpus; amber bar = that target; colored bar = model prediction. At convergence the bars match.

Add corpus words to see constraints.