5- What are possible extensions to more complex robustness scenarios?
5.1- allow for a same trace to map to distinct root-causes (that would require the agent to increase confidence on fix decisions while no becoming stuck on suboptimal ones /explore&exploit process (see issue #62)
5.2- allow for utility to non-stationary, hence the agent would have to decide whether a perturbation was on the trace-fix prediction or on the component-utility prediction (see issue #66)
5.3- allow for asynchronous failures (see corresponding issue #61)
5.4- allow for multiple failures on the same shop (see issue #65)
5.5- allow for perfect failure masking (see issue #64)
5- What are possible extensions to more complex robustness scenarios? 5.1- allow for a same trace to map to distinct root-causes (that would require the agent to increase confidence on fix decisions while no becoming stuck on suboptimal ones /explore&exploit process (see issue #62) 5.2- allow for utility to non-stationary, hence the agent would have to decide whether a perturbation was on the trace-fix prediction or on the component-utility prediction (see issue #66) 5.3- allow for asynchronous failures (see corresponding issue #61) 5.4- allow for multiple failures on the same shop (see issue #65) 5.5- allow for perfect failure masking (see issue #64)