Function extract_unique_forecasts

tsExperiments/scripts_plot/plotting.py:23–50

Extract unique forecasts and their probabilities. Args: samples: Array of shape (n_hyp, forecast_length, target_dim) for TimeGrad or (n_hyp, forecast_length, target_dim+1) for timeMCL is_mcl: Whether the samples come from timeMCL model (with scores)

(samples, is_mcl=True)

#%%

def extract_unique_forecasts(samples, is_mcl=True):
    """Extract unique forecasts and their probabilities.
    
    Args:
        samples: Array of shape (n_hyp, forecast_length, target_dim) for TimeGrad
                or (n_hyp, forecast_length, target_dim+1) for timeMCL
        is_mcl: Whether the samples come from timeMCL model (with scores)
    """
    n_hyp = samples.shape[0]

    reshaped_forecasts = samples.reshape(n_hyp, -1)
    unique_forecasts, unique_indices = np.unique(reshaped_forecasts, axis=0, return_index=True)
    hypothesis_forecasts = samples[unique_indices]
    
    if is_mcl:
        # Use provided scores for probabilities
        probabilities = np.zeros_like(hypothesis_forecasts)
        for i in range(hypothesis_forecasts.shape[0]):
            probabilities[i] = 0
            for j in range(samples.shape[0]):
                if np.all(samples[j] == hypothesis_forecasts[i]):
                    probabilities[i] += 1
        probabilities = probabilities / probabilities.sum(axis=0, keepdims=True)
    else:
        # For TimeGrad, use equal probabilities
        probabilities = np.ones_like(hypothesis_forecasts) / probabilities.sum(axis=0, keepdims=True)

    return hypothesis_forecasts, probabilities

def plot_mcl(target_df, hypothesis_forecasts, forecast_length, rows=4, cols=4, plot_mean=True, context_points=None, freq_type='H', fname='Predictions_plot.png', extract_unique=True, save_path=None, is_mcl=True, plot_p=True, main_color='lightcoral', mean_color='blue', dataset=None, axs=None, ax_cbar=None, global_min_prob=None, global_max_prob=None, dims_to_plot=None, seed=None):
    """