Unsupervised Learning of Temporal Abstractions With Slot-Based Transformers.

The discovery of reusable subroutines simplifies decision making and planning in complex reinforcement learning problems. Previous approaches propose to learn such temporal abstractions in an unsupervised fashion through observing state-action trajectories gathered from executing a policy. However, a current limitation is that they process each trajectory in an entirely sequential manner, which prevents them from revising earlier decisions about subroutine boundary points in light of new incoming information. In this work, we propose slot-based transformer for temporal abstraction (SloTTAr), a fully parallel approach that integrates sequence processing transformers with a slot attention module to discover subroutines in an unsupervised fashion while leveraging adaptive computation for learning about the number of such subroutines solely based on their empirical distribution. We demonstrate how SloTTAr is capable of outperforming strong baselines in terms of boundary point discovery, even for sequences containing variable amounts of subroutines, while being up to seven times faster to train on existing benchmarks.

Furry protein suppresses nuclear localization of yes-associated protein (YAP) by activating NDR kinase and binding to YAP.

The Hippo signaling pathway suppresses cell proliferation and tumorigenesis. In the canonical Hippo pathway, large tumor suppressor kinases 1/2 (LATS1/2) phosphorylate the transcriptional coactivator yes-associated protein (YAP) and thereby suppress its nuclear localization and co-transcriptional activity. Nuclear Dbf2-related kinases 1/2 (NDR1/2), which are closely related to LATS1/2, also phosphorylate and inactivate YAP by suppressing its nuclear localization. Furry (FRY) is a cytoplasmic protein that associates with NDR1/2 and activates them, but its role in the nuclear/cytoplasmic localization of YAP remains unknown. Here, we constructed FRY-knockout cell lines to examine the role of FRY in YAP's cytoplasmic localization. FRY depletion markedly increased YAP nuclear localization and decreased NDR1/2 kinase activity and YAP phosphorylation levels, but did not affect LATS1/2 kinase activity. This indicated that FRY suppresses YAP's nuclear localization by promoting its phosphorylation via NDR1/2 activation. NDR1/2 depletion also promoted YAP nuclear localization, but depletion of both FRY and NDR1/2 increased the number of cells with YAP nuclear localization more strongly than did depletion of NDR1/2 alone, suggesting that FRY suppresses YAP nuclear localization by a mechanism in addition to NDR1/2 activation. Co-precipitation assays revealed that Fry uses its N-terminal 1-2400-amino-acid-long region to bind to YAP. Expression of full-length FRY or its 1-2400 N-terminal fragment restored YAP cytoplasmic localization in FRY-knockout cells. Taken together, these results suggest that FRY plays a crucial role in YAP cytoplasmic retention by promoting YAP phosphorylation via NDR1/2 kinase activation and by binding to YAP, leading to its cytoplasmic sequestration.