Projects

FAQ

Please always consult the FAQ before posting questions to the faculty hosts.

1. Exercise statement, example 1.1.1: expected utility of task T4 is NA.
Output of statement case T01: "state={T1=3.00,T2=1.00,T3=1.00,T4=NA} gain=11.00".
Use the provided testing cases to test differences on your output.

2. Provisory input-output cases: T03 input of version 2 is flawed (there should be TIKs where the agent does not observe utilities).
Still, the reference T03 output is "state={T0=2.00,T1=NA,T2=NA,T3=8.00,T4=0.00,T5=5.50,T6=NA,T7=NA,T8=NA,T9=NA} gain=29.00"

3. T15 output was updated in Mooshak. Click here to see the correct output of T15. Be aware of double precision problems in Java (e.g. 4.999999).

4. In T08, please do not worry with the rounding problems associated with 0.625 and 0.375. The different answers (0.63, 0.62, 0.37 and 0.38) will be counted as corrected to the final evaluation.

5. There are two formats for the input parameter agents: "agents={...}" and "agents=[...]". Only the later, "agents=[...]" will be tested.

6. On simple cases: T05_output: "A1={T1=1.50,T2=2.50}},A2" -> "A1={T1=1.50,T2=2.50},A2"

7. Agents with concurrency penalties. The observed utility (e.g. "A1 u=3") is in fact the gathered utility by the agent. To simplify, assume the agents do not know whether the remaining agents are performing tasks concurrently. As such, agents do not update this perceived utility (e.g. "u=3") according to whether or not there are other agents executing the task. Example: "A1 u=3", if another agent A2 is doing the same task as A1, A1 still believes that the last utility for executing that task is 3.

8. There is another accepted output for test T19: "state={A1={T0=2.00,T1=4.53,T2=NA,T3=NA,T4=2.00,T5=NA},A2={T0=4.83,T1=NA,T2=NA,T3=NA,T4=2.98,T5=NA}} gain=45.00\n" (alternative T19_input_v2 and T19_output_v2)Please note T08 (FAQ entry 4) and T19 tests have more than one valid output. Even if Mooshak gives you wrong answer, these alternative outputs will not be discounted in the final evaluation.

9. A group of agents observes u=3 and u=1 for T1 on time point 1 and u=1 for T1 on time point 2. Should agents sum all utilities OR compute expectations per time point and then sum? Both are conceptually correct and will be accepted. T17 alternative output: state={A1={T0=4.33,T1=6.00,T2=NA,T3=NA,T4=4.00,T5=NA},A2={T0=4.33,T1=6.00,T2=NA,T3=NA,T4=4.00,T5=NA}} gain=49.00

10. To access all alternative outputs allowed in Mooshak, please download input-output cases v4.
Note that the output for the alternative T19: "...A2={T0=4.83,.. gain=45.00\n" and T19b: "...A2={T0=4.83,.. gain=45.00\n"

11. Insight: an agent with memory factor of 0.3 executes T1 in second and third TIK. Expected value of T1 is w_t2 * u_t2 + w_t3 * u_t3 where weight w_t2 = 2^0.3 / (2^0.3+3^0.3).

12. Small differences due to rounding problems (e.g. a difference of 0.01) will not be discounted on the final evaluation. Please note that this is the case with T08, T_alternative_08 or T15, do not worry if they do not succeed in Mooshak.