plot_hmm()

plot_hmm() visualizes fitted HMM parameters: transition matrix, emission matrix, initial probabilities, or a seqHMM-style network graph with pie-chart nodes.

Function Usage

plot_hmm(
    model,
    which="transition",
    figsize=None,
    ax=None,
    vertex_size=50,
    vertex_label_dist=1.5,
    edge_curved=0.5,
    edge_label_cex=0.8,
    vertex_label="initial.probs",
    loops=False,
    trim=1e-15,
    combine_slices=0.05,
    with_legend="bottom",
    layout="horizontal",
    **kwargs
)

seqHMM Parameter Mapping

Sequenzo	seqHMM plot.hmm()
model	Fitted hmm object
which="network"	Default graph view in R
which="transition" / "emission" / "initial"	Matrix-style parameter plots
vertex_size, edge_curved, loops, trim, combine_slices, with_legend, layout	Network layout controls (R plot.hmm arguments)

Entry Parameters

Parameter	Required	Type	Description
model	✓	HMM	Fitted HMM object.
which	✗	str	"transition", "emission", "initial", "network", or "all". Default "transition".
figsize	✗	tuple / None	Matplotlib figure size (width, height).
ax	✗	Axes / None	Existing axes to draw on.
vertex_size	✗	float	Node size for which="network".
vertex_label_dist	✗	float	Label offset for network nodes.
edge_curved	✗	bool / float	Curved transition arrows in network plot.
edge_label_cex	✗	float	Font scale for edge labels.
vertex_label	✗	str	"initial.probs", "names", or custom labels.
loops	✗	bool	Draw self-loop transitions.
trim	✗	float	Hide transitions below this probability.
combine_slices	✗	float	Combine small emission slices into "others" in pie nodes.
with_legend	✗	bool / str	Legend placement: True, False, or "bottom" / "top" / "left" / "right".
layout	✗	str	"horizontal" or "vertical" network layout.
**kwargs	✗	any	Extra arguments passed to the network plot.

What It Returns

A matplotlib Figure object. Call plt.show() in scripts, or display in a notebook.

Example

from sequenzo.seqhmm import build_hmm, fit_model, plot_hmm
import matplotlib.pyplot as plt

hmm = build_hmm(seq, n_states=4, random_state=42)
hmm = fit_model(hmm)

# seqHMM-style network graph
plot_hmm(hmm, which="network", vertex_size=50, edge_curved=0.5)
plt.show()

# All matrix panels
plot_hmm(hmm, which="all")
plt.show()

R Counterpart

• Closest R function: seqHMM plot.hmm()
• Mapping note: Network view mirrors R's default parameter graph; matrix heatmaps are Sequenzo extensions via which.

Notes

• Best used after fitting so plotted values reflect estimated parameters.
• Network plots require optional networkx for layout; matrix plots work without it.
• For mixture models, use plot_mhmm().

Authors

Code: Yuqi Liang

Documentation: Yuqi Liang

References

Helske, S., & Helske, J. (2019). Mixture hidden Markov models for sequence data: The seqHMM package in R. Journal of Statistical Software, 88(3), 1–32.