C.S Projects

Association Rule Mining.

I learnt about Apriori’s theorem, that if a item set is frequent then its subsets must also be frequent. It works by finding the items which appear more and combining into larger and larger item sets as long as the appear frequent enough. This is also an example of unsupervised learning. I think a real life example to use Apriori is when making packages for example in a shop, the items that are often bought together can be made into packages for a cheaper price to promote selling of all the products instead of just one. We also had a quiz with materials from the first week.

I learnt the types of machine learning, which are Supervised, Unsupervised, and reinforcement learning. Supervised allows you to improve performance based on the previous experience and Unsupervised finds all the patterns in the data passed in. Lastly Reinforcement rewards a good choice. I also learnt about the methods of machine learning, like clustering as a technique for unsupervised learning using K- means algorithm which aims to segregate groups with similar traits and assign them into clusters.

I learnt about types of machine learning: Supervised– learns from previous experience, collects the data and produce from that same data. Reinforcement -reward for a correct decision . Unsupervised – finds unknown patterns in the given data . I found it interesting that genetic algorithm and reinforcement learning work on the same principles, that they are inspired by nature and they work on finding the ‘good solution’ according to some pre-defined concept of ‘goodness’. I also learnt to represent belief using probabilities, Bayes Rule to calculate the probabilities.

AI in Games, Adversary Search

I learnt how to represent games as a search problem, defining the starting states, rules, goal test, etc on games like othello, tic tac toe, using mini-max algorithm, alpha beta pruning when the branches of the search tree need not be considered. Also to deal with time limit on adversary search ,iterative deepening search is useful.

Learnt about differences between complete search and partial searching using examples like the TSP, how complete search doesn’t give the optimal solution but rather guarantees that if the goal exists ,the algorithm will reach the goal . The difference between local search and global search, that local search only consider the solutions in the neighborhood while global search considers all possible solutions , I also learnt genetic algorithms use natural selection.