What does the weight suggest?
Move the weight measurement through the five sensor models. The likelihood compares how well each material explains that same reading.
Weight models
p(weight | category), in g⁻¹
Likelihood at this reading
Raw density values; these need not sum to 1.
Normalized likelihood across weight
Dividing by the sum makes a distribution. It is a posterior only under a uniform prior.
A detector reading is evidence.
The vision sensor reports bottle, cardboard, or paper—even when the object is metal. Select its output to compare the likelihood of each category.
Likelihood of the selected detection
The detector model
Each row is p(detection | category) and sums to 100%.
Three sensors. One posterior.
Set the prior and all three measurements. MAP chooses the category with the largest posterior; maximum likelihood uses only the sensor evidence.
From prior to posterior
Outlined bars: prior. Filled bars: posterior. Both share a 0–100% scale.
Follow the evidence
Multiply across each row, then normalize across categories. Weight likelihood is a density (g⁻¹), so the product is not itself a probability.
Watch the posterior change with weight.
Fix conductivity and detection, then sweep across the weight sensor. These curves use the notebook’s prior: 20%, 30%, 25%, 20%, and 5%.