Python代写|COMP30024 Artificial Intelligence


In this second part of the project, we will play the full two-player version of Cachex. Before you read this specification you should re-read the ‘Rules for the Game of Cachex ’ document. The rules of the game are the same as before, however we have further clarified some points of confusion raised by students over the past couple of weeks.

The aims for Project Part B are for you and your project partner to (1) practice applying the gameplaying techniques discussed in lectures and tutorials, (2) develop your own strategies for playing Cachex,and (3) conduct your own research into more advanced algorithmic game-playing techniques; all for the purpose of creating the best Cachex–playing program the world has ever seen.

Your task is twofold. Firstly, you will design and implement a program to play the game of Cachex.

That is, given information about the evolving state of the game, your program will decide on an action to take on each of its turns (we provide a driver program to coordinate a game of Cachex between two such programs so that you can focus on implementing the game-playing strategy). Section 2 describes this programming task in detail, including information about how our driver program will communicate with your player program and how you can run the driver program.

Secondly, you will write a report discussing the strategies your program uses to play the game, the algorithms you have implemented, and other techniques you have used in your work, highlighting the most impressive aspects. Section 3 describes the intended structure of this document.

The rest of this specification covers administrative information about the project. For assessment criteria,see Section 4. For submission and deadline information see Section 5. Please seek our help if you have any questions about this project.

You must create a program in the form of a Python 3.6 module named with your team name.

2.1 The Player class

When imported, your module must define a class named Player with at least the following three methods:

The action must be represented based on the instructions for representing actions in the next section.

The parameter player will be the player whose turn just ended (either “red” or “blue”), and action will be the action performed by that player. Of course, if it was your turn that just ended, action will be the same action you returned through the action method. Again, actions will be represented following the instructions for representing actions in the next section. The action will always be valid since the referee performs validation before this method is called (e.g., your turn method does not need to validate the action against the game rules).

2.2 Representing actions

Our programs will need a consistent representation for actions. We will represent all actions as tuples containing first a string action type and then possible action arguments, indexing hexes with the axial coordinate system from Part A (see Figure 1).

(“STEAL”, )

There are no arguments required. As per the game rules this action may only be played by Blue as their first move of the game.

(“PLACE”, r, q)

The arguments r, q denote the coordinate of the token being placed on the game board. Note that there is no need for a separate capture action since captures occur as a consequence of placing tokens.

2.3 Running your program

To play a game of Cachex with your program, we provide a driver program—a Python module called referee. For your information, the referee program has the following essential structure:

(a) Ask the active player for their next action (calling their .action() method).

(b) Validate the action and apply it to the game if is allowed (otherwise, end the game with an error message). Display the resulting game state to the user.

(c) Notify both players of the action (calling their .turn() methods).

(d) Switch the active player to facilitate turn-taking.

To play a game using referee, invoke it as follows. The referee module (the directory referee/) and the modules with your Player class(es) should be within your current directory:

python -m referee <n> <red module> <blue module>where python is the name of a Python 3.6 interpreter1 , <n> is the size of the game board and <red module>and <blue module> are the names of modules containing the classes Player to be used for Red and Blue, respectively. The referee offers many additional options. To read about them, run ‘pythonm referee –help’.

2.4 Game board size

Cachex may theoretically be played on a hex board of arbitrary size, denoted by the parameter n. However,for practical purposes we will constrain testable n values to the range [3, 15] (inclusive). In fact, the referee program will not accept values outside of this range. While your agent must be capable of playing a game for any n inside this range, you should first prioritise optimising it for n = 8 and n = 9. The majority of assessed test cases will use these board sizes.

2.5 Program constraints

The following resource limits will be strictly enforced on your program during testing. This is to prevent your programs from gaining an unfair advantage just by using more memory and/or computation time. These limits apply to each player for an entire game. In particular, they do not apply to each turn separately. For help measuring or limiting your program’s resource usage, see the referee’s additional options (–help).

You must not attempt to circumvent these constraints. For example, do not use multiple threads or attempt to communicate with other programs or the internet to access additional resources.

2.6 Allowed libraries

Your program should use only standard Python libraries, plus the optional third-party libraries NumPy and SciPy (these are the only libraries installed on dimefox). With acknowledgement, you may also include code from the AIMA textbook’s Python library, where it is compatible with Python 3.6 and the above limited dependencies. Beyond these, your program should not require any other libraries in order to play a game.

However, to develop your program, you may use any tools, including tools using other programming languages. This is all allowed as long as your Player class does not require these tools to be available when it plays a game (because they are not available on dimefox).

For example, let’s say you want to use machine learning techniques to improve your program. You could use third-party Python libraries such as scikit-learn/TensorFlow/PyTorch to build and train a model. You could then export the learned parameters of your model. Finally, you would have to (re)implement the prediction component of the model yourself, using only Python/NumPy/SciPy. Note that this final step is typically simpler than implementing the training algorithm, but may still be a significant task.

Finally, you must discuss the strategic and algorithmic aspects of your game-playing program and the techniques you have applied in a separate file called report.pdf.

This report is your opportunity to highlight your application of techniques discussed in class and beyond,and to demonstrate the most impressive aspects of your project work.

3.1 Report structure

You may choose any high-level structure of your report. Aim to present your work in a logical way, using sections with clear titles separating different topics of discussion.

Here are some suggestions for topics you might like to include in your report. Note: Not all of these topics or questions will be applicable to your project, depending on your approach. That’s completely normal. You should focus on the topics which make sense for you and your work. Also, if you have other topics to discuss beyond those listed here, feel free to include them.

Example questions: What search algorithm have you chosen, and why? Have you made any modifications to an existing algorithm? What are the features of your evaluation function, and what are their strategic motivations? If you have applied machine learning, how does this fit into your overall approach? What learning methodology have you followed, and why? (Note that it is not essential to use machine learning to design a strong player)

Example questions: How have you judged your program’s performance? Have you compared multiple programs based on different approaches, and, if so, how have you selected which is the most effective?

Examples: algorithmic optimisations, specialised data structures, any other significant efficiency optimisations, alternative or enhanced algorithms beyond those discussed in class, or any other significant ideas you have incorporated from your independent research.

Examples: developing additional programs or tools to help you understand the game or your program’s behaviour, or scripts or modifications to the provided driver program to help you more thoroughly compare different versions of your program or strategy.

You should focus on making your writing succinct and getting the level of detail right. The appropriate length for your report will depend on the extent of your work and so aiming for succinct writing will be more appropriate than aiming for a specific word or page count.

For example, there’s probably no need to present detailed code inside your report. Moreover, there’s no need to re-explain ideas we have discussed in class (and if you have applied a technique or idea that you think we may not be familiar with, then it would be appropriate to write a brief summary of the idea and provide a reference through which we can obtain more information).

3.2 Report constraints

While the structure and contents of your report are flexible, your report must satisfy the following constraints: