Reference no: EM133665126
Intelligent Systems
Assignment and Project - AI product
Artificial Intelligence (AI) techniques such as rule-based expert systems, machine learning, fuzzy systems, evolutionary computations, intelligent agents, etc. have emerged as an option for many applications and products. Imagine if you are running an AI startup company, you are to pitch and design your first AI product for the company to get funding for your company. There are two stages to get your AI product funded. The first stage is for you to pitch your prototype to potential investors to get some seed funding to develop the prototype into a product. In the second stage, you are to demonstrate your AI product to the investors and get further funding to fund the mass production and sales of your product. You are required to provide the detail report and the AI product for this project.
AI techniques are being applied in many products in many domains such as games, medical, business intelligence, banking, engineering design, scheduling, data analysis, mining, petroleum, power engineering, weather forecasting, climate prediction, etc. Marks will be awarded for showing the understanding of the problem, the quality of the solution and the steps in building the AI system, the clarity of the report, quality of any code written, the use of tools and the evaluation method. In this assignment and project, you are to build (not just described) the AI techniques for your product.
If you are using any data or APIs that are not publicly available, you should have sorted any legal arrangement and implication. If you are unsure, always discuss with your tutor or unit coordinator.
You will need to provide an investigation of at least two AI techniques and provide a justification on why the selected AI technique is better to be used for your product. For example, if you are designing an obstacle avoidance system for a robot/intelligent agent, you can compare the fuzzy system and rule based system and choose one for your final product development. For another example, if you are designing a real time object recognition system, you can compare the use of AlexNet and YOLO networks. Justification needs to be provided for your choice. To justify your selected AI technique, you need to demonstrate you understand everything of the AI model including the steps of establishing the models. Do not just use it blindly.
In topic 2, you have learned about the phases of developing an Intelligent System as follows.
In the Assignment submission, you will need to include everything from Phase 1 to Phase 3. For Project submission, you will need to submit everything you have done from Phase 1 to Phase 6.
Phase 1: Problem assessment Assignment & Project
Phase 2: Data and knowledge acquisition Assignment & Project
Phase 3: Development of a prototype system Assignment & Project
Phase 4: Development of a complete system Project
Phase 5: Evaluation and revision of the system Project
Phase 6: Integration and maintenance of the system (provide the plan for this project) Project
In IT project development, creating prototype/s is an important step before the final product is developed and deployed. The assignment will require you to submit a prototype of the system/sub- systems and a report describing the development of the prototype. Remember, this is only a prototype, not the full product.
Assignment Report and Code - Submission Requirements
A written formal technical report should cover the following:-
Title of your AI Product (E.g. "An AI Pest Recognition System for Western Australia")
Description of the AI product and how the product can be innovative. Give as much background information about the problem domain you are working on so that others can understand the innovation of your product. Discuss what methods are commonly used and how AI techniques can help to improve the process. (NOTE: You will need to research the topic and propose what AI system can be developed). Provide an understanding of the data or knowledge required for designing such an AI product/solution.
When investigating the suitable AI technique or solution for your suggested AI product, you can cover the following (but not limited to)
How the given information/ knowledge be encoded or learned
Where can you obtain the data/information/knowledge from
Any data pre-processing or data cleaning required
Any successful application that has been reported before
How different is your proposed product from a competitive product
Which AI techniques will be chosen for comparison and why - need to provide critical analysis of at least two AI techniques and provide justification on why the one is selected - if you like you can show some initial test results to justify your choice.
How the output of the system will be interpreted
Any information you think can help the potential investors to understand your innovation and uniqueness of your product. Remember, you need to provide a "product result and not an academic result".
References and any additional information need to be cited. For example, if you use a website on the tutorial of how to create a real time object detection system, you need to provide a reference on this. Although you learn from the website, you need to understand every step as an interview may be required to check if you understand the system you are developing. If you cannot demonstrate that you understand what you are submitting, you could receive a zero for your submission.
When creating the AI product, you will need to discuss and demonstrate an understanding of the following in the final project:-
Why the particular AI technique is chosen for the problem
Provide a detail description of the steps in creating the AI system
Describe and explain any data pre-processing and post-processing steps. Give reasons to justify why they are required for the different AI technique used.
Explain any parameters available in the AI technique that you can optimise or trial, even though if you could be just using the default values. You will need to demonstrate that you understand them.
Why the kind of evaluation method you have chosen is suitable for evaluating your AI system (including the discussion of why such measurement is used)
Justify your investigation and comparison results to recommend the best solution for your problem domain.
Provide an integration and maintenance plan for your product to be used in real world situation.
Note that this project does not require you to provide results that are better than those in the literature or website. However, you need to demonstrate that you understand every step and justify them. You will need to discuss why such results could be acceptable for your product.
A report should be organised as a scientific report under the following headings. Note: The following is only a suggestion, you are allowed to include more details to justify and explain your solution.
Title: A line stating clearly the purpose of the AI product.
Abstract: One paragraph giving a summary of the product, the innovation and uniqueness of the product, the results and conclusions of your work.
Introduction: Describe and discuss the application and the problem that your product can be used for. You need to demonstrate you understand the problem domain you are working on and why your AI product could help to improve the process or workflow.
Background: Describe what prior experiments have been done on such a problem. Put the results of your reading to work. Briefly critique these works and say what you learn from each. Explain how you get the motivation or idea of the proposed AI product. Explain how this leads to your choice of the AI technique, as detailed in the next section.
AI Technique: Give details of the algorithms, heuristics, you are using. Provide a detail description of any parameters you can use to optimise the AI technique. When necessary, give
citations for the sources or research papers for these. Credit any other authors who you have cited, and make sure to be clear about what are your work. Provide the steps in creating the chosen AI techniques. Provide any listings of your code (or models) in the Appendix at the end. Please take note: You have to submit the code or model, and datasets such that your lecturer can repeat your steps here. If you use any (extra) libraries/frameworks to build your solution, it has to be clearly stated in the User Guide (not here).
Evaluation Method. Explain how you will show that you have met the goal of your product. This might involve an analysis of the behaviour of your system's operation, architecture, task and environment. Systematically manipulate one of these variables at a time and see how it influences the behaviour of your system. Alternatively, it might be important to measure the performance of your system using some direct score, measurement or index. Describe the data cases or specific environment you used and the steps you took to set up your experiment. Explain the measurements or observations you will use.
Results: Give the results obtained from your experiments. If a classification task, for example, show how many of a test set were correctly classified, while if it was an optimisation method, show your optimal measure and compare it with a theoretical ideal optimum. If a learning method, you should also include some indication of how quickly, and how well your system learned. This will generally be in the form of a graph, or table of data. Make sure you clearly label all graphs and tables and provide a caption for each. For some applications like games or process operation, it may be easier to show the results in the form of pictures (do screen captures and insert them as with captions "Figure 1. Splash screen showing game environment", "Figure 2. Bot team in play" etc. so that you can refer to them in the text)
Discussion: Interpret the results for the reader. How can we understand what has happened? Was the outcome what you expected? What have you learned by doing the experiment? What went well, and what went poorly? What could be done to improve the performance of the method in future? What is your conclusion?
Acknowledgements (if needed): If you are given help or use code taken from some source, you may want to acknowledge these here. If you modify code obtained from the book or Internet, remember to observe the copyrights of the code obtained, and make proper reference to the author. If it's a requirement for you to gain permission from the author of the code, please make sure you keep the copy of approval obtained and submit a copy with your report.