Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
CSCI 343 Exam 2
| Term | Definition |
|---|---|
| Explain the primary difference between supervised and unsupervised learning, specifically regarding the data used | Supervised learning uses labelled data, while unsupervised learning uses unlabeled data |
| When would you choose to use a regression model over a classification model? | When your labels are numerical (usually continuous) values instead of categorical values |
| in a simple linear regression model ˆy = θ0 + θ1x what is the parameter of θ0 | θ0 is the y intercept |
| in a simple linear regression model ˆy = θ0 + θ1x what is the parameter of θ1x | θ1x is the slope (or weight) associated with x |
| What is the main advantage of using Root Mean Squared Error (RMSE) as a loss function compared to Mean Squared Error (MSE) | RMSE is more interpretable since it is in the same units as the y (label) data |
| Compare Mean Absolute Error (MAE) to Mean Squared Error (MSE). How does their sensitivity to large errors differ? | Since MSE squares the error, it is more sensitive to large errors than MAE |
| Why is Adjusted R^2 often preferred over R^2 when evaluating a multiple linear regression model? | Adjusted R^2 takes into account the number of features, and penalizes the addition of features that do not improve the model's performance more than random change |
| In Adjusted R^2, what does variable n represent? | n is the number of data points |
| In Adjusted R^2, what does variable p represent? | p is the number of features |
| Give an example of when it is a good idea to use a hexbin plot | When you have too many data points for the graph to be readable if you plotted them individually. |
| Give an example of when you may need to use log scale | When plotting large values of money (income/salary, property value, loan amount, etc.) |
| What is the key difference between a histogram and a bar plot? | Histograms are used for quantitative data while bar plots are used for categorical data |
| Why should stacked plots be avoided? | They are hard to read and understand |
| What is the purpose of exploring data? | To understand patterns, detect problems, and prepare data for analysis modeling |
| What are some common issues spotted using summary statistics? | Missing values, invalid values or outliers, data ranges that are too wide/narrow, and inconsistent units |
| What is the goal of data visualization? | To communicate data insights clearly and effectively with minimal cognitive strain for the viewer. |
| What do you need to keep in mind when creating data visualizations? | 1. as much information as possible 2. make data stand out and maintain clarity 3. avoid clutter & too many elements 4. Choose proper aspect ratios and scaling 5. Center data properly; avoid skewing 1 side |
| What are three iterative stages of data visualization? | 1. Graph the data 2. Learn 3. Regraph to answer new questions |
| What types of graphs can be used for single-variable distributions? | Pie charts, histograms, density plots, and bar charts |
| When should you use a pie chart? | For showing proportions that sum to 100%, with few categories (ideally < or equal to 5) |
| Why are bar charts often preferred to pie charts? | They are easier to compare visually, especially when there are more categories |
| What does a histogram show? | The frequency of data within fixed-width bins or intervals. |
| Example of histogram use | Grouping customers' gas bills into $10 intervals to count frequency. |
| What is a density plot? | A smoothed, continuous version of a histogram, where the area under the curve = 1. |
| What does a point on a density plot represent? | A fraction of data that takes on that value |
| What's most important in a density plot? | The shape of the curve, not exact y-values |
| When should you use a logarithmic scale? | When percent change or order of magnitude is more meaningful than absolute values, or when data is heavily skewed. |
| What is a bar chart used for? | displaying frequencies of categorical data or discrete variables |
| Can a bar chart start above zero? | Sometimes, but it depends on context and the story being told |
| What are 4 questions data visualizations can help answer? | 1. What is the peak value? 2. How many peaks? (uni/bi modal) 3. How much variation exists? 4. Is the data concentrated in certain intervals/categories? |
| Which plots work best for 2 continuous variables? | Line graphs, scatter plots, hexbin plots |
| Which plots are best for 2 discrete variables? | Stacked bar charts or side by side graphs |
| What makes line graphs effective? | Each x-value should correspond to a unique y-value; no clutter or excess lines (limit to 3-4) |
| What is an area chart? | A line graph with color shading below the line to show magnitude(mix between bar and line chart) |
| What is a hexbin plot? | A 2D histogram that shows data density using color or shading |
| Why might stacked bar charts be difficult to read? | Because comparing values across stacks is visually harder than comparing side-by-side bars |
| What's a good alternative to stacked bar charts? | Side-by-side charts or multiple small plots |
| What is a model? | An idealized representation of a system, such as weather forecast |
| What is ML used for? | Generating decision-making and predictive models |
| What are the main stages of a ML project? | 1. Problem Definition 2. Data Collection 3. Data Preparation & reprocessing 4. Model Building 5. Evaluation 6. Deployment |
| What are the 3 main types of ML algorithms? | Supervised, Unsupervised, and Reinforcement learning |
| What is supervised learning? | Learning from labeled data to predict outputs (classification or regression) |
| What is unsupervised learning? | Learning from unlabeled data to find hidden patterns or clusters |
| What is reinforcement learning? | Learning to make sequential decisions through trial and error in an environment |
| What is the goal of supervised learning? | To find relationships between inputs (features) and outputs (targets) |
| What does a regression model predict? | Continuous numerical outcomes |
| What does a classification model predict? | Categorical outcomes (high-risk/low-risk, tumor type, dog breed) |
| Predicting Rainfall- what model would be used? | Regression (continuous output) |
| Predicting Gender from an image - what model would be used? | Classification (categorical) |
| Predicting house prices- what model would be used? | Regression |
| What type of learning does linear regression use? | Supervised learning |
| What does linear regression do? | Estimates relationships between a dependent variable and one or more independent variables |
| What is the equation for simple linear regression? | 𝑦^=𝜃0+𝜃1𝑥y^=θ0 +θ1x or 𝑦=𝑚𝑥+𝑏y=mx+b |
| Why are estimates imperfect? | Because data has variability and noise - models simplify reality |
| How does covariance measure? | How two variables vary together |
| What does positive covariance indicate? | When one variable increases, the other tends to increase |
| What does negative covariance indicate? | When one variable increases, the other tends to decrease |
| What does zero covariance indicate? | No consistent relationship between the variables |
| Why is covariance hard to interpret | Its value depends on the units of the variables |
| What is correlation? | a standardized form of covariance that measures the strength and direction of a linear relationship. |
| What range do correlation values fall within? | Between -1 (perfect negative) and +1 (perfect positive) |
| Does correlation imply causation | No! Dur! |
| What is the adjusted R^2 value? | A version of R^2 adjusted for the number of features used; it only increases if a new feature improves the model by chance. |
| What is RMSE (Root Mean Square Error)? | A measure of model prediction error- the square root of the average squared differences between predicted and actual values |
| What does standard deviation represent? | The average amount by which values differ from the mean; a measure of spread |
| What type of task does logistic regression model do? | Classification -- predicting categorical outcomes |
| Give an example of a logistic regression use case | Diagnosing whether a tumor is benign or malignant, predicting COVID test results |
| What kind of outputs does logistic regression produce? | Probabilities that map to categorical classes |
| What is the purpose of data visualization in data analysis? | To explore, understand, and communicate insights from data clearly. |
| what is the main goal of linear regression? | To model and predict a continuous dependent variable based on one or more independent variables. |
| What is the goal of logistic regression? | To model classification tasks where the outcome is categorical, such as "yes/no" or "positive/negative" |
| How is logistic regression different from linear regression? | Linear regression predicts continuous outputs; logistic regression predicts categorical outcomes by estimating probabilities between 0 and 1 |
| What is the range of logistic regression output values? | Between 0 and 1 (probability values) |
| What function converts real-values outputs into probabilities? | the logistic (sigmoid) function |
| What is the formula for the sigmoid function? | σ(x)=1/(1+e^−x1) |
| What does the sigmoid function do in logistic regression? | Maps any real number to a value between 0 and 1, representing probability |
| What type of function is the sigmoid in machine learning? | an activation function |
| How can you classify data using a regression model? | Draw a cutoff (threshold). Values above it are classified as one class, values below it as another. |
| What is a classification threshold? | The probability value used to decide which class a data point belongs to (commonly .5) |
| What loss function is used in binary logistic regression? | Binary Cross Entropy (BCE) Loss |
| What is the goal of BCE Loss? | To measure how well predicted probabilities match actual binary outcomes. |
| What activation function is paired with BCE Loss? | The sigmoid function |
| What loss function is used in multinomial logistic regression? | Categorical Cross Entropy (CCE) Loss |
| What activation function is used for multiclass logistic regression? | The softmax function |
| What does the softmax function do? | converts a vector of raw scores into probabilities that sum to 1 across all classes |
| How is the predicted class label determined in multiclass classification? | By the class with the highest probability from the softmax output |
| What is a confusion matrix? | a table comparing predicted vs actual outcomes, showing counts of true/false positives and negatives |
| What are the four basic terms in a confusion matrix? | 1. TP (True Positive): Model correctly predicts positive class. 2. FP (False Positive): Model incorrectly predicts positive 3. TN (True Negative): Model correctly predicts negative class 4. FN (False Negative): Model incorrectly predicts negative |
| What does precision measure? | The proportion of correctly predicted positive samples out of all predicted positives |
| Precision formula | TP/ (TP + FP) |
| Why is high precision important? | When false positives are costly, like in medical diagnoses |
| What does Recall (sensitivity) measure? | The proportion of actual positives that were correctly identified |
| Recall formula | TP/(TP+FN) |
| Why is high recall important? | When missing a positive case is costly, such as fraud detection or disease screening |
| What is Specificity? | The proportion of actual negatives that were correctly identified. |
| Specificity Formula | TN/(TN+FP) |
| What is False Positive Rate? (FPR) | The proportion of negative samples incorrectly predicted as positive |
| False Positive Rate formula | FP/(FP + TN) |
| What is False Negative Rate? (FNR) | The proportion of positive samples incorrectly predicted as negative. |
| False Negative Rate formula | FN/(FN+TP) |
| What is Accuracy? | The overall percentage of correctly classified samples |
| Accuracy Formula | (TP+TN)/(TP+TN+FP+FN) |
| When is accuracy misleading? | When classes are imbalanced (99% negatives, 1% positives) |
| What is the F1 Score? | The harmonic mean of precision and recall |
| F1 Formula | 2 * (precision * recall)/(precision + recall) |
| Why use the F1 score? | It balances false positives and false negatives, especially useful with imbalanced datasets |
| What is a ROC curve? | A plot of the True Positive Rate (Recall) against the False Positive Rate at various thresholds |
| What does the Area Under the Curve (AUC) represent? | The model's ability to distinguish between classes |
| What does AUC =1 mean? | Perfect Classifier |
| What does AUC =.5 mean? | No discriminative ability (random guessing) |
| What does a higher AUC indicate? | Better classification performance. |
| Why should training and testing data be separate? | To prevent overfitting- the model must be evaluated on unseen data to assess generalizaiton |
| Why can summary metrics be misleading? | They may hide issues like poor data preprocessing or implementation errors |
| What is the best way to evaluate model performance? | Use unseen (test) data and multiple metrics, not just accuracy) |
| What is the difference between training, validation, and test sets? | Training data teaches the model; validation tunes it; test data evaluates final performance |
| What is multiple linear regression? | A model that uses two or more features to predict a target. |
| Multiple Linear Regression Formula | y^=θ0+θ1x1+θ2x2+...+θnxn |
| What does adding more features to a model do? | Can improve accuracy if features are useful, but risks overfitting |
| What is a loss function? | A formula that quantifies how far predictions are from actual values |
| List common loss functions for regression | MAE,MSE,RMSE |
| Define Mean Absolute Error (MAE) | Average of absolute differences |yi-^yi| (less sensitive to outliers) |
| Define Mean Squared Error (MSE) | Average of squared differences (yi-^yi)^2; penalizes large errors more |
| Define Root Mean Squared Error (RMSE) | Square root of MSE; interpretable in same units as target variable |
| What is the goal when fitting a model? | Minimize the average loss on training data |
| What is R^2 | The coefficient of determination; proportion of variance in y explained by x; best score = 1, can be negative if model is poor |
| How does logistic regression produce probabilities? | Applied sigmoid function to a linear equation to output values in [0,1] |
Created by:
user-1979725