COSC 3337 Exam One with verified solutions A+ rated

COSC 3337 Exam One with verified
solutions A+ rated

Supervised Learning - correct answer ✔✔- Using data to predict an outcome



Unsupervised Learning - correct answer ✔✔- Using data to group items/users into categories



Reinforcement Learning - correct answer ✔✔- Optimizing action based on response variable



Supervised Learning Algorithms - correct answer ✔✔- Linear Regression/Logistic Regression

- Random Forest/Decision Trees/Gradient Boosting

- SVM

- Maximum Likelihood

- Neutral Nets



Linear Regression - correct answer ✔✔Trying to find a function that will minimize the distance in
between the data



Unsupervised Learning Algorithms - correct answer ✔✔- K-Means Clustering

- Decision Tree Clustering

- Topic Models

- SVM

- Gaussian Mixture Models



Topic Models - correct answer ✔✔- Cluster content into a finite collection of "topics"



Reinforcement Learning Algorithms - correct answer ✔✔- Relies entirely on maximizing expectation of
reward conditioned on user attributes and action

,- Incredibly useful since it's actionable

- Easiest when you can measure variable without major outside interference



5 Steps for Approaching an Application - correct answer ✔✔- Define the problem to be solved

- Collect data

- Choose an algorithm class

- Choose an omptimization metric for learning the model

- Choose a metric for evaluating the model



Eager Learner - correct answer ✔✔- Begins classifying as it receives data set

- Does not wait for test data to learn

- Takes a long time learning and less time classifying



Lazy Learner - correct answer ✔✔- Stores data set without learning from it

- Starts classifying as it receives test data

- Takes less time learning and more time classifying



Online Learning - correct answer ✔✔- Based on each pattern as it is observed



Batch Learning - correct answer ✔✔- Learns over groups of patterns. Most algorithms are based off of
this.



Parametric Algorithm - correct answer ✔✔- Fixed number of parameters

- Computationally faster, but makes stronger assumptions about the data

- Linear Regressionq!~



Nonparametric Algorithm - correct answer ✔✔- Uses a flexible number of params and the number
grows as it learns from more data

- Computationally slowe

, - KNN



Discriminative Models - correct answer ✔✔- Try to only understand how classes are separated based on
attributes



Generative Models - correct answer ✔✔- Try to understand probability distribution of the data (x,y)



Machine Learning - correct answer ✔✔- Small Training Dataset

- Chosen Features

- Large number of algorithms

- Short training time



Deep Learning - correct answer ✔✔- Large training dataset

- No chosen features

- Small number of algorithms

- Long training time



Attribute - correct answer ✔✔- Property or characteristic of an object

- Variable, field, characteristic

- EX: Eye color, temperature, etc.



Nominal Attributes - correct answer ✔✔- Variables that provide descriptive information about an object
(city name, vegetation type, color). No order



Ordinal Attributes - correct answer ✔✔- Have a ranking or order to their data (ranking from 1 to 10, or
high, med, low, etc, white, lt grey, grey, dark grey, black).



Interval Attributes - correct answer ✔✔- Used for date, temperature, time

- Zero is arbitrary, doesn't mean absence. (e.g. 0 degrees = equator not absence of latitude)