|  dehnert | a7caf709ae | default to topological equation solver | 7 years ago | 
				
					
						|  dehnert | dff67450e0 | fixed recently introduced bug in JANI export | 7 years ago | 
				
					
						|  dehnert | 8114437cee | allowing cumulative and instantaneous reward properties to be transformed to JANI | 7 years ago | 
				
					
						|  dehnert | d638972bc8 | enabled pushing location assignments to edges | 7 years ago | 
				
					
						|  TimQu | 749ba87254 | Made SVI log output more clear | 7 years ago | 
				
					
						|  Sebastian Junges | 98969e627c | updated counterexamples to support statistics to be exported | 7 years ago | 
				
					
						|  dehnert | db6f43ed9d | made LRA computation for deterministic systems able to respect that the underlying solver requires a fixed-point formulation | 7 years ago | 
				
					
						|  dehnert | 86069b8552 | fix typo in JSON exporter | 7 years ago | 
				
					
						|  dehnert | 50aa6d1424 | assuming the only global real transient variable is the reward when exporting JANI and no reward model is mentioned in the property (issues a warning) | 7 years ago | 
				
					
						|  dehnert | c4bed85dc4 | switching to native linear equation solver by default and power iteration | 7 years ago | 
				
					
						|  TimQu | 1f6fc7e273 | Better conversion of MA to CTMC if there are only Markovian states | 7 years ago | 
				
					
						|  TimQu | ca2295be1d | updated changelog: support for expected total rewards | 7 years ago | 
				
					
						|  TimQu | 5a16b2befa | minor fixes to let the total reward tests compile and pass | 7 years ago | 
				
					
						|  Matthias Volk | 081c0a95d0 | Export pnpro with single-server semantics | 7 years ago | 
				
					
						|  TimQu | 1f4c0325be | test cases for ctmcs and markov automata | 7 years ago | 
				
					
						|  TimQu | 8df9b461cb | total reward formulas for ctmcs and markov automata | 7 years ago | 
				
					
						|  TimQu | b5566fa861 | more on total reward formulas for mdps | 7 years ago | 
				
					
						|  TimQu | b3edae8707 | fixed fragment specification: total reward formulas should not be supported for hybrid/dd right now | 7 years ago | 
				
					
						|  Alexander Bork | 8c3bd15eae | Fixed priorities for dependencies and export of PDEP probabilities into PNPRO format | 7 years ago | 
				
					
						|  Matthias Volk | 7dc17065c1 | Updated DFT export to new JSON format | 7 years ago | 
				
					
						|  Sebastian Junges | 0be0126095 | fixed support for highlevel counterex for expected rewards in dtmcs | 7 years ago | 
				
					
						|  Sebastian Junges | 73a1911a53 | Merge branch 'master' into counterexample_improvements | 7 years ago | 
				
					
						|  Alexander Bork | 1850cf7368 | Fixed trigger rates for timed transitions not being saved in .pnpro files | 7 years ago | 
				
					
						|  TimQu | c2dd57cda5 | total rewards for mdps | 7 years ago | 
				
					
						|  TimQu | 87e34d7b32 | Added Support for Total Reward Formulas for DTMCs in the Sparse Engine | 7 years ago | 
				
					
						|  dehnert | dfc0141894 | minor fix to Z3 API modification | 7 years ago | 
				
					
						|  dehnert | cdfa328464 | first attempt at adapting to Z3 interface change | 7 years ago | 
				
					
						|  Matthias Volk | 1f221db280 | Disable transformation of DFT properties to JANI | 7 years ago | 
				
					
						|  Matthias Volk | eea940b625 | Refactoring for transformation DFT->GSPN->JANI | 7 years ago | 
				
					
						|  dehnert | 4134630fa6 | adding gap output | 7 years ago | 
				
					
						|  Alexander Bork | 758382e020 | Merge remote-tracking branch 'origin/dft_gspn' into dft_gspn | 7 years ago | 
				
					
						|  Alexander Bork | a49f88b7f5 | Fixed computation of priorities to correctly represent the semantics | 7 years ago | 
				
					
						|  dehnert | 4f4d1f4423 | do not scale precision | 7 years ago | 
				
					
						|  Matthias Volk | a6e6d5993f | Travis: set unlimited clone depth to allow versioning with git describe | 7 years ago | 
				
					
						|  Matthias Volk | 0d4cf67f2e | Set mergeDC from setting | 7 years ago | 
				
					
						|  Matthias Volk | cf316df35e | Added settings for DFT-GSPN transformation | 7 years ago | 
				
					
						|  Matthias Volk | afb0be1245 | Fixed missing dependencies to storm-parsers | 7 years ago | 
				
					
						|  Matthias Volk | 9e398ffaab | Minor improvements for some CMake output | 7 years ago | 
				
					
						|  dehnert | 386f0b2e47 | fixing bug in scheduler improvement step of policy iteration for games | 7 years ago | 
				
					
						|  Matthias Volk | 369cea775d | Swapped order in PriorityQueue to propagate failures in correct order | 7 years ago | 
				
					
						|  sjunges | 8478352030 | dynamic constraints and minimality labels | 8 years ago | 
				
					
						|  dehnert | 2aff2e9382 | adding some more timing output | 8 years ago | 
				
					
						|  sjunges | 53238f43f7 | fixed some missing includes due to updated API | 8 years ago | 
				
					
						|  sjunges | 39698d6ecb | fix install of storm-counterexamples | 8 years ago | 
				
					
						|  sjunges | 79bb6734ed | compile and link parsers in seperate binary | 8 years ago | 
				
					
						|  sjunges | 3a704ae532 | fix storm-dft missing includes | 8 years ago | 
				
					
						|  sjunges | 6dfce6a405 | extended counterexamples towards expected rewards, and moved counterexamples to a seperate lib (still in main cli) to slightly accelarate building times | 8 years ago | 
				
					
						|  dehnert | b0047a5a96 | improving choices in game policy iteration depending on precision of underlying solver | 8 years ago | 
				
					
						|  sjunges | 6fcc91b9d0 | Merge branch 'master' of https://srv-i2.informatik.rwth-aachen.de/scm/git/storm | 8 years ago | 
				
					
						|  Sebastian Junges | d1ab260068 | remove outdated parts in highlevel counterexamplse | 8 years ago |