This problem is finding the shortest path from a given start state while collecting the necessary dots, or like in Pacman sense, "food pallets".
We deploy these methods to solve the problem:
- Depth-first search
- Breadth-first search
- A* search
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp1/assignment1.html
(Part 2 couldn't be done due to time constraints)
This problem transforms a 2D planning problem for a robotic arm into a configuration space, and then search for a path in that space.
We deploy these methods to solve the problem:
- Calculate the configuration space map
- Use the bfs search algorithm created in MP1 to make it work
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp2/assignment2.html
This problem trains teh biniary senitment classifier with a datset of movie reviews to distinguish between positive and negative reviews of the movie.
We deploy these methods to solve the problem:
- Naive Bayes algorithm to train a binary sentiment
- Using Bayes Theorem to compute the probability of a revieww being positive/negative
- Using MAP classification to predict which review sets are positive
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp2/assignment3.html
This problem implements part of sppech (POS) tagging using an HMM model. It first trains the data with tags and uses it to test data without tags.
We deploy these methods to solve the problem:
- HMM trellis (Viterbi) decoding algorithm
- Employ techniques to maximize the viterbi algorithm
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp2/assignment4.html
This problem uses perceptron to teach a computer to distingusih livign things from non-living things. More precisely, to distinguish if an image contains animals or not.
We deploy these methods to solve the problem:
- A simple perceptron algorithm to detect an image
- A simple model to decide whether or not images after using perceptron classifier to train the computer
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp2/assignment5.html
This problem uses more complex perceptron to tackle the image problem presented using MP5.
We deploy these methods to solve the problem:
- Using nerual networks (multilayer perceptron) to train the image
- Evaluate the value Fw by using classical sigmoid network to determine the images
- Imporve performance by using various techniques (such as choosing activiation function, reularization, and etc.)
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp2/assignment6.html
This problem creates RL agent to balance pole.
We deploy these methods to solve the problem:
- Q-learning in the tabular setting and function approximation setting
Further details about the MP task can be seen in the link below: https://courses.grainger.illinois.edu/cs440/fa2019/MPs/mp2/assignment7.html