ava.plotting package

Submodules

ava.plotting.grid_plot module

Plot a grid of spectrograms.

ava.plotting.grid_plot.grid_plot(specs, gap=3, vmin=0.0, vmax=1.0, ax=None, save_and_close=True, filename='temp.pdf')

Parameters:

• specs (numpy.ndarray) – Spectrograms
• gap (int or tuple of two ints, optional) – The vertical and horizontal gap between images, in pixels. Defaults to 3.
• vmin (float, optional) – Passed to matplotlib.pyplot.imshow. Defaults to 0.0.
• vmax (float, optional) – Passed to matplotlib.pyplot.imshow. Defaults to 1.0.
• ax (matplotlib.pyplot.axis, optional) – Axis to plot figure. Defaults to matplotlib.pyplot.gca().
• save_and_close (bool, optional) – Whether to save and close after plotting. Defaults to True.
• filename (str, optional) – Save the image here.

ava.plotting.grid_plot.indexed_grid_plot_DC(dc, indices, ax=None, save_and_close=True, gap=3, side_len=128, filename='grid.pdf')

Plot a grid of spectrograms.

Parameters:

• dc (ava.data.data_container.DataContainer) – Data container object.
• indices (list of lists of ints) – Spectrogram indices.
• ax (matplotlib.axes._subplots.AxesSubplot or None) – Plotting axis. Defaults to None.
• save_and_close (bool) – Save and close the plot. Defaults to True.
• gap (int) – Number of pixels between spectrograms. Defaults to 3.
• side_len (int) – Spectrogram height and width, in pixels. Defaults to 128.
• filename (str) – Save the image here. Defaults to ‘grid.pdf’.

ava.plotting.latent_projection module

Plot a latent mean projection.

ava.plotting.latent_projection.latent_projection_plot_DC(dc, embedding_type='latent_mean_umap', color_by=None, title=None, filename='latent.pdf', colorbar=False, colormap='viridis', alpha=0.5, s=0.9, ax=None, cax=None, shuffle=True, save_and_close=True, show_axis=False, default_color='b', condition_func=None)

Make a scatterplot of latent means projected to two dimensions.

Parameters:

• dc (ava.data.data_container.DataContainer) – See ava.data.data_container.
• embedding_type (str, optional) – Defaults to 'latent_mean_umap'.
• color_by ({str, None}, optional) – If None, all points are colored default_color. Otherwise, color_by is requested from the DataContainer and passed to the matplotlib.pyplot.scatter as the color parameter. The one exception is if color_by == 'filename_lambda', in which case, scatter color is some function of the audio filename, passed to condition_func. Defaults to None.
• title ({str, None}, optional) – Plot title. Defaults to None.
• filename (str, optional) – Where to save the image, relative to dc.plots_dir. Defaults to 'latent.pdf'.
• colorbar (bool, optional) – Whether to include a colorbar. Defaults to False.
• colormap (str, optional) – The pyplot colormap to use if color_by returns numerical values. Defaults to 'viridis'.
• alpha (float, optional) – Alpha value of scatterpoints. Defaults to 0.5.
• s (float, optional) – Size of scatterpoints. Defaults to 0.9.
• ax ({matplotlib.axes._subplots.AxesSubplot, None}, optional) – Scatter axis. If None, matplotlib.pyplot.gca() is used. Defaults to None.
• cax ({matplotlib.axes._subplots.AxesSubplot, None}, optional) – Colorbar axis. If None, an axis is made. Defaults to None.
• shuffle (bool, optional) – Whether to shuffle to zorder of points. Defaults to True.
• save_and_close (bool, optional) – Defaults to True.
• show_axis (bool, optional) – Defaults to False.
• default_color (str, optional) – Defaults to 'b'.
• condition_func ({function, None}, optional) – Only used when color_by == 'filename_lambda', in which case condition_func maps audio filenames to pyplot colors. Defaults to None.

ava.plotting.latent_projection.latent_projection_plot_with_noise_DC(dc, noise_box, embedding_type='latent_mean_umap', color_by=None, title=None, filename='latent.pdf', colorbar=False, colormap='viridis', alpha=0.5, s=0.9, ax=None, cax=None, shuffle=True, save_and_close=True, show_axis=False, default_color='b', condition_func=None)

Same as latent_projection_plot_DC, but with noise to exclude.

Parameters:

• dc (ava.data.data_container.DataContainer) – See ava.data.data_container.
• noise_box (list of float) – Must contain four elements: [x1, x2, y1, y2], which are interpreted as a region of the latent mean UMAP embedding containing noise. The points within this rectangle are excluded.

Note

For other parameters, see latent_projection_plot_DC.

ava.plotting.latent_projection.projection_plot(embedding, color='b', title=None, save_filename='latent.pdf', colorbar=False, shuffle=True, colormap='viridis', alpha=0.6, s=0.9, ax=None, cax=None, save_and_close=True, show_axis=False)

Parameters:

• embedding (numpy.ndarray) –

  …

• color ({str, numpy.ndarray}, optional) – Defaults to 'b'.
• title ({str, None}, optional) – Defaults to None.
• save_filename (str, optional) – Defaults to 'temp.pdf'.
• colorbar (bool, optional) –

  …

• shuffle (bool, optional) –

  …

ava.plotting.mmd_plots module

Maximum Mean Discrepancy plots.

Note

• Because MMD^2 is the value that is directly estimated, this is the value that is saved and passed between functions. In v0.3.0, variable names have been changed to make this distinction clear.

[1]	Gretton, Arthur, et al. “A kernel two-sample test.” Journal of Machine Learning Research 13. Mar (2012): 723-773. http://www.jmlr.org/papers/volume13/gretton12a/gretton12a.pdf

ava.plotting.mmd_plots.estimate_median_sigma(latent, n=10000, seed=42)

Estimate the median pairwise distance for use as a kernel bandwidth.

Parameters:

• latent (numpy.ndarray) – Latent means.
• n (int, optional) – Number of random pairs to draw. Defaults to 10000.
• seed ({None, int}) – Random seed. Defaults to 42.

Returns:

sigma – Median pairwise Euclidean distance between sampled latent means.

Return type:

float

ava.plotting.mmd_plots.mmd_matrix_plot(mmd2, cluster=True, cmap='viridis', ax=None, colorbar=True, cax=None, ticks=[0.0, 0.3], filename='mmd_matrix.pdf', save_and_close=True)

Plot a pairwise MMD matrix.

Parameters:

• mmd2 (numpy.ndarray) – Pairwise MMD^2 values, a square matrix.
• cluster (bool, optional) – Whether to order conditions by a clustering algorithm. Defaults to True.
• cmap (str, optional) – Name of matplotlib colormap. Defaults to 'viridis'.
• ax (matplotlib.axes._subplots.AxesSubplot, optional) – Matplotlib axis. Defaults to the current axis, plt.gca().
• colorbar (bool, optional) – Whether to plot a colorbar. Defaults to True.
• cax (matplotlib.axes._subplots.AxesSubplot or None, optional) – Colorbar axis. If None, a new axis is made. Defaults to None.
• ticks (list of floats, optional) – Colorbar ticks. Defaults to [0.0, 0.3].
• filename (str, optional) – Where to save plot. Defaults to 'mmd_matrix.pdf'.
• save_and_close (bool, optional) – Save and close the figure. Defaults to True.

ava.plotting.mmd_plots.mmd_matrix_plot_DC(dc, condition_from_fn, mmd2_fn, condition_fn, parallel=False, load_data=False, cluster=True, alg='quadratic', max_n=None, sigma=None, cmap='Greys', colorbar=True, cax=None, ticks=[0.0, 0.3], filename='mmd_matrix.pdf', ax=None, save_and_close=True)

Plot a pairwise MMD matrix.

Parameters:

• dc (ava.data.data_container.DataContainer) – DataContainer object.
• condition_from_fn (function) – Returns an int representing condition, given a filename.
• mmd2_fn (str) – Where MMD^2 values are saved to/loaded from.
• condition_fn (str) – Where conditions are saved to/loaded from.
• parallel (bool, optional) – Whether to calculate different MMD^2 values in parallel. If True, MMD^2 values are printed out to stdout and can then be saved and formed into a proper matrix using the _matrix_from_txt helper function.
• load_data (bool, optional) – Whether to load precomputed data. Defaults to False.
• cluster (bool, optional) – Whether to order conditions by a clustering algorithm. Defaults to True.
• alg ({'linear', 'quadratic'}, optional) – Use the linear-time or quadratic time MMD^2 estimate. Defaults to 'quadratic'.
• max_n (int or None, optional) – Maximum number of samples from each distribution. If None, no maximum is set. Only applies if alg == 'quadratic'. Defaults to None.
• sigma ({float, None}, optional) – Kernel bandwidth. If None, the median distance is used. Defaults to None.
• cmap (str, optional) – Name of matplotlib colormap. Defaults to 'viridis'.
• colorbar (bool, optional) – Whether to plot a colorbar. Defaults to True.
• cax (matplotlib.axes._subplots.AxesSubplot or None, optional) – Colorbar axis. If None, a new axis is made. Defaults to None.
• ticks (list of floats, optional) – Colorbar ticks. Defaults to [0.0, 0.3].
• filename (str, optional) – Where to save plot, relative to dc.plots_dir. Defaults to 'mmd_matrix.pdf'.
• ax (matplotlib.axes._subplots.AxesSubplot, optional) – Matplotlib axis. Defaults to the current axis, plt.gca().
• save_and_close (bool, optional) – Whether to save and close the plot. Defaults to True.

ava.plotting.mmd_plots.mmd_tsne_plot_DC(dc, mmd2_fn=None, condition_fn=None, mmd2=None, conditions=None, perplexity=30.0, s=4.0, alpha=0.8, label_func=None, ax=None, save_and_close=True, filename='mmd_tsne.pdf', load_data=False)

Compute and plot a t-SNE layout from an MMD matrix.

Either pass mmd2 and conditions directly, or specify mmd2_fn and condition_fn and set load_data=True.

• add option to calculate MMD^2.

Parameters:

• dc (ava.data.data_container.DataContainer) – DataContainer object.
• mmd2_fn (str) – Where MMD^2 values are saved to/loaded from.
• condition_fn (str) – Where conditions are saved to/loaded from.
• mmd2 ({numpy.ndarray, None}, optional) – MMD^2 matrix. Defaults to None.
• conditions ({numpy.ndarray, None}, optional) – Condition for each entry of the MMD^2 array. Defaults to None.
• perplexity (float, optional) – Passed to t-SNE. Defaults to 30.0.
• s (float, optional) – Passed to matplotlib.pyplot.scatter. Defaults to 4.0.
• alpha (float, optional) – Passed to matplotlib.pyplot.scatter. Defaults to 0.8.
• label_func ({function, None}, optional) – Maps a conditions to a label (string) for annotating points. Defaults to None.
• ax (matplotlib.axes._subplots.AxesSubplot, optional) – Matplotlib axis. Defaults to the current axis, plt.gca().
• save_and_close (bool, optional) – Save and close the figure. Defaults to True.
• filename (str, optional) – Where to save plot. Defaults to 'mmd_tsne.pdf'.
• load_data (bool, optional) – Whether to load the MMD^2 and condition data from mmd2_fn and condition_fn. Defaults to False.

ava.plotting.tooltip_plot module

Plot a syllable projection with spectrograms appearing as tooltips.

ava.plotting.tooltip_plot.save_image(data, filename)

https://fengl.org/2014/07/09/matplotlib-savefig-without-borderframe/

ava.plotting.tooltip_plot.tooltip_plot(embedding, images, output_dir='temp', num_imgs=10000, title='', n=30000, grid=False)

Create a scatterplot of the embedding with spectrogram tooltips.

• Set the aspect ratio to 1

Parameters:

• embedding (numpy.ndarray) – The scatterplot coordinates. Shape: (num_points, 2)
• images (numpy.ndarray) – A spectrogram image for each scatter point. Shape: (num_points, height, width)
• output_dir (str, optional) – Directory where html and jpegs are written. Deafaults to “temp”.
• num_imgs (int, optional) – Number of points with tooltip images. Defaults to 10000.
• title (str, optional) – Title of plot. Defaults to ‘’.
• n (int, optional) – Total number of scatterpoints to plot. Defaults to 30000.
• grid (bool, optional) – Show x and y grid? Defaults to False.

ava.plotting.tooltip_plot.tooltip_plot_DC(dc, embedding_type='latent_mean_umap', output_dir='html', num_imgs=1000, title='', n=30000, grid=False, img_format='.jpg')

DataContainer version of tooltip_plot.

Parameters:

• dc (ava.data.data_container.DataContainer) – See ava.data.data_container for details.
• embedding_type (str, optional) – Defaults to 'latent_mean_umap'.
• output_dir (str, optional) – Directory where html and jpegs are written. Deafaults to 'temp'.
• num_imgs (int, optional) – Number of points with tooltip images. Defaults to 1000.
• title (str, optional) – Title of plot. Defaults to ''.
• n (int, optional) – Total number of scatterpoints to plot. Defaults to 30000.
• grid (bool, optional) – Show x and y grid? Defaults to False.
• img_format (str, optional) – Filetype for tooltip spectrograms. Defaults to '.jpg'.

ava.plotting.tooltip_plot.write_images(embedding, images, output_dir='temp/', num_imgs=100, n=30000)

Module contents

AVA plotting module

Contains

ava.plotting.grid_plot

Plot grids of spectrograms.

ava.plotting.latent_projection

Plot scatterplots of latent means.

ava.plotting.mmd_plots

Plot MMD matrices.

ava.models.tooltip_plot

Plot an html plot with spectrogram tooltips.