CS 325 T7 L1,2,3

A specification of a mathematical or probabilistic relationship between different variables in a dataset.

It builds algorithms that learn from data and make predictions or decisions based on new unseen data without human intervention.

It involves models trained on labeled data to predict outcomes based on input variables.

It involves models that work on unlabeled data to find hidden patterns or groupings, such as clustering.

It is a type of machine learning where a model learns through trial and error, guided by rewards or penalties.

Producing a model that performs well on training data but poorly on new, unseen data.

Producing a model that does not perform well on training data and does not generalize well to new data.

The difference between the average prediction of the model and the true value, leading to oversimplification and high prediction error.

The variability of model predictions for a given dataset, where high variance leads to poor performance on new data.

The process of selecting a subset of relevant features from the original dataset to improve model efficiency and accuracy.

The process of deriving new features from the original data to enhance the model's performance.

By minimizing bias and variance to reduce total prediction errors.

A dataset used to train and teach the model, typically 70% of the input dataset.

A dataset used to evaluate the model’s performance, typically 30% of the input dataset.

To build a model that learns from historical data and makes accurate predictions or decisions on new data.

It reduces model complexity, enhances computational efficiency, and reduces generalization errors introduced by noise.

A confusion matrix is a method used to evaluate the performance of a classification model by measuring its correctness.

True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN).

TP is when the model correctly predicts a positive class, like predicting a dog, and the actual outcome is also a dog.

FP is when the model predicts a positive class incorrectly, like predicting a dog when the actual outcome is a cat. This is also called Type I error.

FN is when the model fails to predict a positive class, like predicting not a dog when the actual outcome is a dog. This is also called Type II error.

Accuracy is the proportion of correct predictions out of the total predictions, calculated as (TP + TN) / (TP + TN + FP + FN).

Precision is the proportion of true positives out of all predicted positives, calculated as TP / (TP + FP).

Sensitivity (or recall) is the proportion of actual positives that were correctly predicted, calculated as TP / (TP + FN).

The F1 score is a balanced measure of accuracy, calculated as 2TP / (2TP + FP + FN), harmonizing precision and recall.

The k-NN algorithm is based on the assumption that similar things exist in close proximity, classifying data based on its nearest neighbors.

The main disadvantage of k-NN is its high computational cost, especially when calculating distances between all training data points.

The curse of dimensionality refers to the exponential increase in modeling error as the number of features or dimensions in the data increases, making the model lose predictive power.

Dimensionality reduction is a technique that helps minimize the curse of dimensionality by reducing the number of features used in modeling.

If "k" is too low (like k=1 or k=3), the model can be noisy due to the influence of outliers.

If "k" is too high, it can lead to computational difficulties and loss of model effectiveness.

The preferred value for "k" is usually 5, as it's a good balance between model accuracy and computational efficiency.

To predict categorical outcomes based on conditional probability.

It assumes that all features are independent of each other.

It is based on conditional probability.

Filtering junk email.

Word frequencies or word count.

It processes very fast and can easily be trained on small datasets.

It assumes all variables are independent, which can oversimplify relationships.

It means that given random variables X, Y, and Z, X is independent of Y given Z.