Repos » blog.jread.com » content » posts » how-useful-was-my-ai-short-course-from-2019 » part2.md
/ I have a huge amount of notes from this module, most of which get pretty technical because I built on top of it and did a lot of my own research. The module talked considerably about data, structured vs unstructured, data spaces, data quality and access. It went into detail about Supervised learning, and I've probably used that area of learning mostly to understand what different types of models are good for.
- Supervised Learning Regression - typically used for numerical analysis, like predicting house prices, or predicting trend lines. Classification - of data into categories, such as assigning people a credit rating, ot splitting a complex scatter plot into x number of categories. ** Both of these types of learning are used by make predictions, and algorithms include; Linear regressions, logistic regression, Decision trees, Random Decision Forest, LASSO and ride regression.
- Unsupervised Learning Clustering (eg K Means clustering) - eg: color quantization in images Pattern detection / anomaly detection - eg: netflix recommendations
- Reinforcement learning ** Learns by examining the outcomes that follow each behavior, an AI models needs to determine which series of "moves" would maximize the positive outcome, regardless of whether feedback from the environment is immediate or delayed.
- Deep Learning This is where there is so much interest in the "modern" field of AI, where models are build around neural networks. CNN = Convolutional Neural Networks rely on weights on nodes. ** RNN = Recurrent neural networks - same, but includes loops