AWS Certification Prep - General Terms

  

AWS Certification Prep

General AI

General Misc Abbreviations

APIs = Application Programming Interfaces

docs = documents

ex = example

GPU = Graphics Processing Unit

IDE = Integrated Development Environment

ISV = Independent Software Vendor

NIST = National Institute of Standards and Technology

OWASP = Open Web Application Security Project

PII = Personally Identifiable Info

SME = Subject matter expert.

General Terms

Bias = data missing important feature due to not sampling enough or correct data. Types are selection/sampling bias (unrepresentative data), interaction/participant bias (user do not want to tell so prejudiced), reporting/measurement bias (over-represents one group), and confirmation/automation bias (reinforcing existing stereotypes, when output is over trusted), and historical/societal bias (from data given in a historical setting tracking data that we feel important).

Decision trees = CASE statement algorithm. Good for transparency.

Explainability = explains WHY after we get the results. Black box. Good with debugging and troubleshooting. 

Interpretability = explains HOW it might make decisions, before it does anything. White box. Good with accountability and builds trust. transparency of model. 

Latency = Time to get response to request.  Real-time applications require low latency.

Nondeterminism = you can run it with the same input data, and it produces different outputs.

Overfitting = when works on the training data but not on the actual evaluation data. Because the model is memorizing the data it has seen and fails to generalize to unseen examples.

Underfitting = when works on the training data but not on the actual evaluation data. Because the model is not capturing all the features of the data and fails to generalize to unseen examples.

Variance = sensitivity to noise or overfitting.

General Math Abbreviations

AUC = Area Under the Curve

GLM = Generalized Linear Models

KNN = K-Nearest Neighbors

LDA = Latent Dirichlet Allocation algorithm 

MSE = Mean Squared Error

PCA = Principal Component Analysis algorithm 

RMSE = Root Mean Squared Error

ROC = Receiver Operator Curve

TF-IDF = Term Frequency-Inverse Document Frequency

XGBoost = eXtreme Gradient Boosting

Math Terms

"Accuracy" in Classification = successes / total 

Classification  = data divided into categories. 

"Dimension" in ML = dataset property or characteristic expressed as a math dimension in vector math. Almost the same as "feature". ex = 3rd number in array.

F1 Score in Classification = harmonic mean of precision and recall. F1 = 2 x (Precision x Recall) / (Precision + Recall).

Generalized Linear Models = number of algorithms that allow linear regression with non-normal data.

Example regressions are: Logistic, Possion, Gamma, and Tweddie.

Linear regression = algorithm to find "best-fit line" of min diff between actual and model's predictions. Disadv: does not work with non-normal data.

Precision in Classification = total positives / total positives + false negatives

Recall in Classification = total positives / total true + false positives

Regularization = penalizes extreme weight values to help prevent linear models from overfitting.

Root Mean Squared Error = standard metric for regressions. calcs avg difference between predict and actual then square roots it.

Vectors = numerical N-dimensional arrays. Ex: Y and X coordinates if 2-D array.

Vector Space = the cloud shape of all the points of all of the vectors.

VAEs = encoder and decoder where latent space follows a probability distribution such as Gaussian.

XGBoost = gradient boosted trees algorithm. Good at regression, classifying, and ranking. Good on tabular data.