14 Questions
What is the main difference between a question-driven and data-driven approach?
The approach to formulating a research question
What is the correct order of the steps to set up and execute an experiment within data science?
Task definition, data collection, data exploration, pre-processing, model learning, evaluation
What is an example of a challenge of working with data?
Noisy data
What does a clear definition of a task based on a given data set and general problem description consist of?
Research question, whether the data is supervised or unsupervised, and data type
What is the equation for simple linear regression?
Y = βX + b
What is the purpose of normalization/standardization in multiple linear regression?
To compare different ranges
What is the role of cross-entropy loss in logistic regression?
To measure the difference between predicted and actual probabilities
What is the purpose of the sigmoid function in logistic regression?
To transform the output to a probability distribution
What is the purpose of the sigmoid function in the given context?
To transform the output into a probability
What type of data would an image be classified as?
Unstructured data
What is the purpose of gradient descent in the given context?
To minimize the loss function
What is the advantage of reporting the median instead of the mean when the data is skewed?
It is more informative
What type of data would a database with rows and columns be classified as?
Structured data
What is data exploration often a precursor to?
Machine learning
Study Notes
Data Science Research Approaches
- Question-driven approach: formulate hypothesis, design experiment, collect data, analyze data, accept or reject hypothesis
- Data-driven approach: explore data, formulate research question, structure and annotate data, develop and apply learning techniques, evaluate on data, answer research question
Steps in Data Science Research
- Formulate a clear research question and task definition
- Collect data
- Explore and preprocess data
- Develop and apply learning techniques
- Evaluate model performance
- Answer research question
Challenges of Working with Data
- Noisy data
- Large data
- Small data
- Incomplete data
- Different sampling rates
- Different formats
- Wrongly chosen or irrelevant variables
- Large or unknown number of classes
- Class imbalance
- Heterogeneous data or features
- New domain
Defining a Task in Data Science
- Clearly define the research question
- Specify whether the task is supervised or unsupervised
- Determine the type of task: classification, regression, ranking, etc.
- Identify the data and its characteristics
- Identify the labels or targets and their characteristics
Simple and Multiple Linear Regression
- Simple linear regression: Y = βX + b, where Y is the dependent variable, X is the independent variable, β is the slope, and b is the intercept
- Multiple linear regression: y = β⋅X + b, requires normalization or standardization of values before training
Logistic Regression
- A discriminative model that learns to distinguish between two classes
- Learns from a training set to minimize the loss function L(y ̂,y)
- Uses gradient descent to optimize the parameters w and b
- Uses the sigmoid function to transform output to a probability: σ(z) = 1/(1+e^(-z) )
Data Types and Exploration
- Structured data: a database with rows, columns, and a relational key
- Semi-structured data: data with additional information, such as references to images
- Unstructured data: unorganized data, such as images, text, or time series data
- Use median instead of mean when data is skewed or has outliers
- Boxplots are more useful for outlier detection than histograms
Learn about the question-driven and data-driven approaches in data science research, including the steps involved in each methodology. Understand how to set up and execute an experiment within data science.
Make Your Own Quizzes and Flashcards
Convert your notes into interactive study material.
Get started for free