From 346093179c23fa50919df349cc39f47a47548a1d Mon Sep 17 00:00:00 2001 From: Andrea Censi Date: Fri, 16 Jun 2023 15:03:00 +0200 Subject: Update README.md --- README.md | 44 +++++++++++++++++++++++++++++++++++++------- 1 file changed, 37 insertions(+), 7 deletions(-) diff --git a/README.md b/README.md index 0fc73b3..92560f6 100644 --- a/README.md +++ b/README.md @@ -1,3 +1,16 @@ +# ACT4EI, Spring 2023, DP exercise + +Note: this code exercise is **optional** -- it is very interesting to do, but we do not feel we have play-tested it enough for it to be a graded exercise. + +## Overview + +There are 2 repositories to consider: + +- This repository contains these instructions and the solution template. +- The repository [ACT4E/ACT4E-mcdp-exercise](https://github.com/ACT4E/ACT4E-mcdp-exercise) contains the test runner and the data structures to use. + Note that there are extensive docs available online. + + # Setup ## Install required software @@ -163,23 +176,40 @@ You should see the output: TODO: finish this part -# ACT4E exercises requirements +--- + +# ACT4E exercises progression + +## Phase 1: solve simple DP queries for each component -## Compulsory +In this first phase, you will implement the queries for each type of DP in isolation. -The following queries must run successfully: +The following query must run successfully: act4e-mcdp-solve-dp-queries -d downloaded/lib1-parts --solver act4e_mcdp_solution.DPSolver + +## Phase 2: solve DP queries with multiple DPs (series, parallel, loop) + +In this phase, you will implement the queries for *composition of DPs* joined by series, parallel, loop constructions. + +The following query must run successfully: + act4e-mcdp-solve-dp-queries -d downloaded/lib2-simple --solver act4e_mcdp_solution.DPSolver -## Optional + +## Phase 3: solve MCDP queries + +In this phase, you are given only the graph. You then need to convert the graph into a DP. + +To do this, you might need some experience or strong intuition about manipulating graphs. The following should work successfully: act4e-mcdp-solve-mcdp-queries -d downloaded/lib1-parts --solver act4e_mcdp_solution.MCDPSolver act4e-mcdp-solve-mcdp-queries -d downloaded/lib2-simple --solver act4e_mcdp_solution.MCDPSolver +## Phase 4: solve advanced queries -## Optional (advanced) +These are advanced queries that require a nontrivial implementation to solve efficiently. - act4e-mcdp-solve-dp-queries -d downloaded/lib3-advanced --solver act4e_mcdp_solution.MCDPSolver - act4e-mcdp-solve-mcdp-queries -d downloaded/lib3-advanced --solver act4e_mcdp_solution.MCDPSolver \ No newline at end of file + act4e-mcdp-solve-dp-queries -d downloaded/lib3-advanced --solver act4e_mcdp_solution.DPSolver + act4e-mcdp-solve-mcdp-queries -d downloaded/lib3-advanced --solver act4e_mcdp_solution.MCDPSolver -- cgit v1.2.1 From bb16b876fb7fad019006c7177325d073763b8e50 Mon Sep 17 00:00:00 2001 From: Andrea Censi Date: Fri, 16 Jun 2023 15:41:17 +0200 Subject: Update README.md --- README.md | 140 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 130 insertions(+), 10 deletions(-) diff --git a/README.md b/README.md index 92560f6..50d4156 100644 --- a/README.md +++ b/README.md @@ -1,19 +1,117 @@ # ACT4EI, Spring 2023, DP exercise -Note: this code exercise is **optional** -- it is very interesting to do, but we do not feel we have play-tested it enough for it to be a graded exercise. +Note: for Spring 2023 this exercise is **optional** -- it is very interesting to do, but we do not feel we have play-tested it enough for it to be a graded exercise. -## Overview + + +## Repositories There are 2 repositories to consider: -- This repository contains these instructions and the solution template. -- The repository [ACT4E/ACT4E-mcdp-exercise](https://github.com/ACT4E/ACT4E-mcdp-exercise) contains the test runner and the data structures to use. - Note that there are extensive docs available online. +- **This repository** contains these instructions and **the solution template** to extend. +- The repository [ACT4E/ACT4E-mcdp](https://github.com/ACT4E/ACT4E-mcdp) contains: + + 1. the API (**data structures**, interfaces); + 2. the **test runner** that checks that your implementation is correct; + 3. the **test cases**. + +Note that there are extensive docs available online at [act4e-mcdp.readthedocs.io](https://act4e-mcdp.readthedocs.io/en/latest/). + +What you'll do is **clone this repository**, and just *install* the code from `ACT4E-mcdp` using `pip` (no need to clone that). + + +## Overview of the exercise + +The goal of this exercise is for you to build a DP and MCDP queries solver. + +There are various phases of difficulty progression (listed at the end of this document). + +You will be given a DP represented using the data structures defined in `ACT4E-mcdp` along with a query. + +You have to extend the code in this repository by implementing the `DPSolver` and `MCDPSolver` classes. + +For example, this is the interface of the `DPSolver` class: + +```python +from act4e_mcdp import DPSolverInterface, Interval, LowerSet, PrimitiveDP, UpperSet + +class DPSolver(DPSolverInterface): + def solve_dp_FixFunMinRes( + self, + dp: PrimitiveDP, + functionality_needed: object, + /, + resolution_optimistic: int = 0, + resolution_pessimistic: int = 0, + ) -> Interval[UpperSet[object]]: + # just a template -- return infeasibility! + optimistic = pessimistic = UpperSet([]) + return Interval(pessimistic=pessimistic, optimistic=optimistic) + + def solve_dp_FixResMaxFun( + self, + dp: PrimitiveDP, + resource_budget: object, + /, + resolution_optimistic: int = 0, + resolution_pessimistic: int = 0, + ) -> Interval[LowerSet[object]]: + + # just a template -- return infeasibility! + optimistic = pessimistic = LowerSet([]) + return Interval(pessimistic=pessimistic, optimistic=optimistic) +``` + +The two methods `solve_dp_FixFunMinRes` and `solve_dp_FixResMaxFun` correspond to the two queries studied in class +(also documented [here](https://act4e-mcdp.readthedocs.io/en/latest/API/API_solution/#act4e_mcdp.solution_interface.DPSolverInterface)). + +The input `dp` represents the DP in consideration. There is a hierarchy of DPs available ([documented here][here]). +These includes the one that appeared in class or in the book (identity, addition, multiplication, splitters, etc.). + +[here]: https://act4e-mcdp.readthedocs.io/en/latest/API/primitivedps/00_index/ + +The inputs (`functionality_needed` and `resource_budget`) are elements of the poset `F` or `R` for the DPs. +For these exercises we have [3 types of posets](https://act4e-mcdp.readthedocs.io/en/latest/API/posets/): + +1. `FinitePoset`: finite posets, with strings as elements +2. `Numbers`: with Python `Decimal` as elements. +3. `PosetProduct`, whose elements are *tuples* (`(a, b, c)`) of elements of its subposets. + +The return value is an `Interval` of upper / lower sets. The two extram of the interval represent optimistic and pessimistic solutions. + +The template code above returns empty upper / lower sets, which means that the solution is not feasible. + +The parameters `resolution_optimistic` and `resolution_pessimistic` are needed in the case the DP is not finitely computable and needs an approximated solution. In that case, your code would be called with increasing values of the resolutions and the expectation is that the interval becomes tighther and tighther. + +### Workflow +This is the suggested workflow: -# Setup +1. Setup everything as described in the next section. +2. Run the test runner and observe the errors. +3. Select one error and work on that. This usually means adding support for a new type of DP or poset. +4. repeat from 2 -## Install required software +**Note**: the `ACT4E-mcdp` package only contains the data structures to describe the problems. It does not contain the actual "code"! For example, the `PosetProduct` class available has a member `subs` that contains the subposets. However, there is no `leq()` or `meet()` method implemented. + + + + +## Turning in the exercise + +You can either email us a `.zip` file, or (better), create a **private** repository and gives us access. + +We ask you that **you do not make your code public**. + + +# Setup the development environment + +This walkthrough makes you setup the development environment on VS Code in a docker container. +It works everywhere including Windows and it does not assume much about your system. + +If you know what you are doing, you could also (at your responsibility and without our support) use a native editor and Python environment. + +## Install Docker Install Docker: @@ -23,6 +121,8 @@ Install Docker: go [here](https://docs.docker.com/desktop/windows/install/) and follow the "Install Docker Desktop on Windows" instructions. You can then start Docker Desktop and follow the quick start quide. +## Install VS Code + Install VS Code [using the instructions online](https://code.visualstudio.com/download). ## Open the folder in VS Code using "Dev Container." @@ -43,7 +143,7 @@ You can create a new terminal using Terminal -> New Terminal. ## Install dependencies -Install this package and dependencies using: +In a terminal (inside the container), install this package and dependencies using: pip install -e . @@ -180,6 +280,12 @@ TODO: finish this part # ACT4E exercises progression +There are 4 phases of increasing difficulty. + +Phase 1 and 2 are approachable, and were designed to be the graded part. + +Phases 3 and 4 are more challenging, and require some extra effort, and were always thought as optional. + ## Phase 1: solve simple DP queries for each component In this first phase, you will implement the queries for each type of DP in isolation. @@ -196,7 +302,7 @@ The following query must run successfully: act4e-mcdp-solve-dp-queries -d downloaded/lib2-simple --solver act4e_mcdp_solution.DPSolver -## Phase 3: solve MCDP queries +## Phase 3: solve MCDP queries (*) In this phase, you are given only the graph. You then need to convert the graph into a DP. @@ -207,9 +313,23 @@ The following should work successfully: act4e-mcdp-solve-mcdp-queries -d downloaded/lib1-parts --solver act4e_mcdp_solution.MCDPSolver act4e-mcdp-solve-mcdp-queries -d downloaded/lib2-simple --solver act4e_mcdp_solution.MCDPSolver -## Phase 4: solve advanced queries +## Phase 4: solve advanced queries (*) These are advanced queries that require a nontrivial implementation to solve efficiently. act4e-mcdp-solve-dp-queries -d downloaded/lib3-advanced --solver act4e_mcdp_solution.DPSolver act4e-mcdp-solve-mcdp-queries -d downloaded/lib3-advanced --solver act4e_mcdp_solution.MCDPSolver + + + + +# Creating more test cases + +The models are created using [this online environment][IDE] - very experimental! + +[IDE]: https://editor.zuper.ai/editor/gh/co-design-models/ACT4E-exercises-spring2023/main/view/ + +If you make an account you can play around. You can obtain the YAML file read by `ACT4E-mcdp` by selecting the "DP compiled YAML representation" option from the drop down menu. Some of the other visualizations might also be helpful. + +[Example: drone model](https://editor.zuper.ai/editor/gh/co-design-models/ACT4E-exercises-spring2023/main/view/libraries/lib2-simple/specs/models/things/drone1/) + -- cgit v1.2.1 From b91f6208d48c0665df56fa851564e3b7930e2d6d Mon Sep 17 00:00:00 2001 From: Andrea Censi Date: Fri, 16 Jun 2023 15:44:34 +0200 Subject: Update README.md --- README.md | 3 +++ 1 file changed, 3 insertions(+) diff --git a/README.md b/README.md index 50d4156..34b29d6 100644 --- a/README.md +++ b/README.md @@ -333,3 +333,6 @@ If you make an account you can play around. You can obtain the YAML file read by [Example: drone model](https://editor.zuper.ai/editor/gh/co-design-models/ACT4E-exercises-spring2023/main/view/libraries/lib2-simple/specs/models/things/drone1/) + mcdpdrone + + -- cgit v1.2.1