ClockInference

clock_inference_engine. ClockInference

This is the main class of the script. It applies and adjusts the KernelDensityEstimation to the clock interfaces of Nomon.

Constructor

new ClockInference(parent, bc, past_data)

Source:
Properties:
Name Type Description
kde KernelDensityEstimation

The instance of the KernelDensityEstimation class.
clocks_li Array.<number>

Array of all the active clocks on the screen.
cscores Array.<number>

Array of scores for the likelihood that the user intended to choose a particular clock at the current round of selection, for all clocks on the screen. This is a list of scores for only the current round of selection.
clock_history Array.<Array.<number>>

2D Array where each sub-array lists the relative angles around each clock at the click times so far. For example, if the user has clicked twice so far, clock_history would be a length-2 Array where each of the two elements is an Array of the relative angles around each clock for all the clocks. Note that the most recent press corresponds to clock history[0].
clock_locs Array.<number>

The relative angles around each clock at the press times only for the most recent press. It is equivalent to clock_history[0].
Parameters:
Name Type Description
parent Keyboard

The instance of the main Keyboard Class.
bc BroderClocks

The instance of the BroderClocks class.
past_data Object | null

Object containing previous data from the KDE from the prior session, if no past data then null. Details in the KernelDensityEstimation class documentation.

Methods

get_score_inc(yin) → {number}

Description:

  • Calculates the likelihood that a clock will be chosen, given the relative click angle around that clock. Note that the input is the click time converted into the relative angle around the clock, not the actual click time itself.

Source:
Parameters:
Name Type Description
yin number

The relative click time in seconds
Returns:
  • returns the likelihood that a clock will be chosen.
Type
number

handicap_cscores(is_win, is_start)

Description:

  • If the highest scoring clock is above a certain threshold, force the highest score to be that threshold.

Source:
Parameters:
Name Type Description
is_win boolean

Was the top-score clock selected in the current round.
is_start boolean

Is this the first round before any clicks.

inc_score_inc(yin)

Description:

  • Note that the density calculation only involves the most recent n hist samples. When a new yin comes in, decide if we need to throw away any old yin in y_list (anything past n_hist), and call increment_dens(). Also update ksigma calculation.

Source:
Parameters:
Name Type Description
yin number

The relative click time in seconds.

is_winner() → {boolean}

Description:

  • Determine if the highest-probability clock is the winner and should be selected. if its posterior probability is higher than 1−P(error) (There is a winner iff this is met). Refer to page 17, section “Selection Decisions” of Prof. Broderick’s Nomon Tech Report for more detail.

Source:
Returns:
  • Whether the top-score clock should be selected.
Type
boolean

learn_scores(is_undo)

Description:

  • While this function has a very similar name with update scores, they serve quite different functionalities. Instead of doing something directly related with scores, this function updates all history-related attributes, such as win_history, clock_history, and calls inc_score_inc().

Source:
Parameters:
Name Type Description
is_undo boolean

Whether the previously selected clock was undo.

reverse_index_gsi(log_dens_val) → {*}

Source:
Parameters:
Name Type Description
log_dens_val
Returns:
Type
*

update_dens(new_time_rotate)

Description:

  • Update the histogram (kde.dens_li) if the rotation period changes. Note that this function is only used when the rotation period changes, and the kde.dens_li when a new press time comes in happens at inc_score_inc().

Source:
Parameters:
Name Type Description
new_time_rotate number

The new clock rotation period in seconds.

update_history(time_diff_in)

Description:

  • Given a new click time, update clock_history.

Source:
Parameters:
Name Type Description
time_diff_in number

The relative click time in seconds.

update_scores(time_diff_in)

Description:

  • Update the posterior estimates for the clocks (cscores) every time a yin comes in.

Source:
Parameters:
Name Type Description
time_diff_in number

The relative click time in seconds.

update_sorted_inds()

Description:

  • Updates the list of clocks sorted by highest cscore (the largest posterior probability).

Source: