English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

This paper proposes a methodology for discovering meaningful properties indata by exploring the latent space of unsupervised deep generative models. Wecombine manipulation of individual latent variables to extreme values outsidethe training range with methods inspired by causal inference into an approachwe call causal disentanglement with extreme values (CDEV) and show that thisapproach yields insights for model interpretability. Using this technique, wecan infer what properties of unknown data the model encodes as meaningful. Weapply the methodology to test what is meaningful in the communication system ofsperm whales, one of the most intriguing and understudied animal communicationsystems. We train a network that has been shown to learn meaningfulrepresentations of speech and test whether we can leverage such unsupervisedlearning to decipher the properties of another vocal communication system forwhich we have no ground truth. The proposed technique suggests that spermwhales encode information using the number of clicks in a sequence, theregularity of their timing, and audio properties such as the spectral mean andthe acoustic regularity of the sequences. Some of these findings are consistentwith existing hypotheses, while others are proposed for the first time. We alsoargue that our models uncover rules that govern the structure of communicationunits in the sperm whale communication system and apply them while generatinginnovative data not shown during training. This paper suggests that aninterpretation of the outputs of deep neural networks with causal methodologycan be a viable strategy for approaching data about which little is known andpresents another case of how deep learning can limit the hypothesis space.Finally, the proposed approach combining latent space manipulation and causalinference can be extended to other architectures and arbitrary datasets.

Approaching an unknown communication system by latent space exploration and causal inference