Reference no: EM133399561

Question 1
Submit your final project. You are expected to submit a document containing the code and the obtained results, and a 2~5 minute video showing how the project runs.

Final Project Document
Include all the project and dataset description that you have submitted in the previous assignment. If you have modified anything, make sure to include the latest descriptions. Add to that the project code. Include both, the code as text, and screenshots of the code. Also, include screenshots that show the obtained results when you run the code.

Project Video
Include a short video (around 2-5 minutes) explaining your project and show how it works. Run the code, enter any necessary input, show the obtained results. Explain what the project is about, what the code does, and what the obtained results mean. You can use your ASU zoom account to record a video, capturing your screen and showing the code running, store it in the cloud, and include the link to the video.

Question 2
Reinforcement Learning
Reinforcement learning is more suitable for simulated environments, and environments in which making mistakes has very low to zero cost. For the applications listed below, list whether reinforcement learning is suitable or not, explaining why. Feel free to include details that better define the environment or the conditions under which reinforcement learning would or would not be suitable.
1. Medical diagnosis of cancer patients
2. Recommending the next action to an auto-pilot
3. Exploring a area with limited space to explore
4. Identifying plants based on their physical features
5. Recommending buying/selling decisions for stocks
6. Controlling a robot arm that assembles toys
7. Training a self-driving car in a simulated environment
8. Training a self-driving car in a real environment (driving in real streets)

Question 3
Discussion: Ethical Issues Related to AI
.AI is being used in many applications related to various life aspects. This led to rising concerns about un-ethical uses of AI.
One controversial example for AI usage is deep fakes. Many have shared concerns about how such technology, powered by AI, can be used in illegitimate ways. You can read more about deep fakes here.
In this discussion, write about one use of AI that you think could be unethical. You can talk about deep fakes, or any other AI technology of your choice.
Grading Criteria:
1. List an AI Application (5 points)
2. Explain what this application is about (10 points)
3. List one usage for this application that may be considered unethical, explaining why that could be the case. (15 points)

Question 4
Training a Room Explorer using RL

Run the script after modifying the main code. Answer the questions below.
1. How does the optimal policy path look like? Draw the path on a 4 X 4 grid.
2. Does it contain as much right or left turns as the policies generated by the old reward function?
3. What is the number of times the exploration task was carried out until the optimal policy was found? (If you are lucky, the code will re run a couple of hundred times. If not, it can keep running for a couple of thousand times before it finds the optimal exploration policy. It should not take more than a couple of minutes maximum.)

Question 5
Using a FSM to program a Self-Driving Car Agent

Keep in mind that there are multiple ways by which a self-driving vehicle can be designed. The choice for a specific method depends on the task environment.
A FSM method would be suitable for simpler environments, with a relatively small number of states, and a small number of input percepts and actions. That's because FMS design relies on listing all percepts at different states, and deciding which action should be taken accordingly. Below is a FSM developed by a research team for high-level decision making in a self-driving car:

Components of a FSM
The finite state machine shows a set of states, representing different actions, and how the transition occurs from one action to another based on the input percept. To design a FSM, you need to first identify:
1. Input Percept: All input percepts, and the values they may take.
2. States: All possible actions/states.
3. Transition Model: How the transition should happen from one state to the another based on the perceived input.

Question 6
Word Suggestion with N-Grams
In this lab, we will generate N-grams for a given text corpus, and use it to make suggestions for the next word to type.

Attachment:- Reinforcement Learning.rar