5.5. pypfilt.obs

All observations models should derive the following base class:

class pypfilt.obs.Obs

The base class of observation models, which defines the minimal set of methods that are required.

log_llhd(ctx, op, time, obs, curr, hist)

Return the log-likelihood \(\mathcal{l}(y_t \mid x_t)\) for the observation \(y_t\) and every particle \(x_t\).

Parameters:
  • ctx – The simulation context.
  • op – The observation model parameters dictionary.
  • time – The current simulation time, \(t\).
  • obs – An observation for the current time-step, \(y_t\).
  • curr – The particle state vectors, \(x_t\).
  • hist – The particle state histories, indexed by observation period.
expect(ctx, op, time, period, prev, curr)

Return the expected observation value \(\mathbb{E}[y_t]\) for every particle \(x_t\), at one or more times \(t\).

Parameters:
  • ctx – The simulation context.
  • op – The observation model parameters dictionary.
  • time – The simulation time(s), \(t\).
  • period – The duration of the observation period (in days).
  • prev – The state vectors at the start of the observation period(s), \(x_t\).
  • curr – The state vectors at the end of the observation period(s).
quantiles(ctx, op, time, mu, wt, probs)

Return the values \(y_i\) that satisfy:

\[y_i = \inf\left\{ y : p_i \le \sum_i w_i \cdot \mathcal{L}(y_t \le y \mid x_t^i)\right\}\]
Parameters:
  • ctx – The simulation context.
  • op – The observation model parameters dictionary.
  • time – The current simulation time, \(t\).
  • mu – The expected case fraction for each particle, \(\mathbb{E}(y_t)\).
  • wt – The weight associated with each particle, \(w_i\).
  • probs – The probabilities \(p_i\), which must be sorted in ascending order.
simulate(ctx, op, time, period, expected, rng=None)

Return a random sample of \(y_t\) for each particle \(x_t\).

Parameters:
  • ctx – The simulation context.
  • op – The observation model parameters dictionary.
  • time – The simulation time(s), \(t\).
  • period – The duration of the observation period (in days).
  • expected – The expected observation value for each particle \(x_t\).
  • rng – The (optional) random number generator to use.
from_file(filename, time_scale)

Load observations from a space-delimited text file with column headers defined in the first line.

Parameters:
  • filename – The file to read.
  • time_scale – The simulation time scale.
Returns:

A list of observations, ordered as per the original file, and the underlying data table.
Return type:

Tuple[List[dict], numpy.ndarray]
pypfilt.obs.expect(ctx, time, unit, period, prev, curr)

Return the expected observation value \(\mathbb{E}[y_t]\) for every every particle \(x_t\), at one or more times \(t\).

Parameters:
  • ctx – The simulation context.
  • time – The simulation time(s).
  • unit (str) – The observation units (see classes in :py:module`data`).
  • period (int) – The duration of the observation period (in days).
  • prev (numpy.ndarray) – The state vectors at the start of the observation period(s).
  • curr (numpy.ndarray) – The state vectors at the end of the observation period(s).
pypfilt.obs.log_llhd_of(ctx, hist, hist_ix, obs, max_back=None)

Return the log-likelihood of obtaining observations from each particle.

Parameters:
  • ctx – The simulation context.
  • hist – The particle history matrix.
  • hist_ix – The index of the current time-step in the history matrix.
  • obs – The observation(s) that have been made.
  • max_back – The number of time-steps into the past when the most recent resampling occurred (i.e., how far back the current particle ordering is guaranteed to persist; default is None, no limit).
Returns:

An array containing the log-likelihood for each particle.
pypfilt.obs.bisect_cdf(probs, cdf_fn, bisect_fn, y_lower, y_upper)

Use a bisection method to estimate the values \(y_i\) that satisfy:

\[y_i = \inf\left\{ y : p_i \le \sum_i w_i \cdot \mathcal{L}(y_t \le y \mid x_t^i)\right\}\]
Parameters:
  • probs – The probabilities \(p_i\), which must be sorted in ascending order.
  • cdf_fn – The CDF function \(f(y) = \sum_i w_i \cdot \mathcal{L}(y_t \le y \mid x_t^i)\).
  • bisect_fn – The bisection function f(a, b) that either returns the midpoint of the interval \([a, b]\) or None if the search should stop (e.g., because a tolerance has been reached).
  • y_lower – A lower bound for all \(y_i\).
  • y_upper – An upper bound for all \(y_i\).
pypfilt.obs.simulate(ctx, time, unit, period, expected, rng=None)

Return a random sample of \(y_t\) for each particle \(x_t\).

Parameters:
  • ctx – The simulation context.
  • time – The simulation time.
  • unit – The observation unit.
  • period – The duration of the observation period (in days).
  • expected – The expected observation value for each particle \(x_t\).
  • rng – The (optional) random number generator to use.