ABSTRACT
Autonomous robots can learn to perform visual navigation tasks from offline human demonstrations and generalize well to online and unseen scenarios within the same environment they have been trained on. It is challenging for these agents to take a step further and robustly generalize to new environments with drastic scenery changes that they have never encountered. Here, we present a method to create robust flight navigation agents that successfully perform vision-based fly-to-target tasks beyond their training environment under drastic distribution shifts. To this end, we designed an imitation learning framework using liquid neural networks, a brain-inspired class of continuous-time neural models that are causal and adapt to changing conditions. We observed that liquid agents learn to distill the task they are given from visual inputs and drop irrelevant features. Thus, their learned navigation skills transferred to new environments. When compared with several other state-of-the-art deep agents, experiments showed that this level of robustness in decision-making is exclusive to liquid networks, both in their differential equation and closed-form representations.
ABSTRACT
BACKGROUND: Candida krusei (Ck) may cause severe infections in immunocompromised hosts and is innately resistant to fluconazole. PATIENTS AND METHODS: During an 18-month period, seven patients with Ck infection were identified at our center. All were treated in the transplant intensive care unit. Candida isolates were grown on Sabouraud agar, and chromosomal DNA was extracted; clonality was investigated using random amplified polymorphic DNA-polymerase chain reaction with primers M13, OPA-18, and OPE-18. RESULTS: Among the patients with Ck infection, there were three pancreas recipients with intra-abdominal infection, one liver recipient with cholangitis, one lung recipient with pleural empyema, one patient with pleural empyema after esophageal perforation, and one case of pneumonia in a patient with a ventricular assist device. Treatment consisted of caspofungin (n = 3), voriconazole (n = 1), or a combination of the two (n = 2) together with surgery (n = 3) or pigtail catheter drainage (n = 3). One patient underwent drainage without antifungal treatment, and one patient did not have drainage. The infection was controlled in all cases. The patient with the assist device died from multiple organ dysfunction, the lung recipient died after four months from graft failure, and one pancreas graft was lost. Four patients (57%) harbored the same Ck strain. CONCLUSION: Solid organ recipients seem to be at particular risk for Ck infections; clonal outbreaks may occur in intensive care units.
