Week 1

Principles of Data Graphics

Data Density

• Data density = Number of entries of data matrix / Available display area

Data Integrity

1. Data scrubbing

• Deliberately removing data or deriving data (smoothing/resampling) to create a false message

Algorithm	Type	Generative/Discriminative	Linear/Non-linear	Notes
Naive Bayes	Classification	Generative	Linear	Assumes feature independence, uses Bayes' theorem
Logistic Regression	Classification	Discriminative	Linear	Uses sigmoid function, mainly for binary classification
Decision Trees	Classification	Discriminative	Non-linear	Handles both numerical and categorical features
SVM	Classification	Discriminative	Linear/Non-linear	Uses kernel trick for non-linear, finds optimal hyperplane
k-NN	Classification	Discriminative	Non-linear	Instance-based, sensitive to feature scaling
Neural Networks	Classification	Discriminative	Non-linear	Can model complex patterns, requires large datasets
Linear Regression	Regression	Discriminative	Linear	Predicts continuous values, sensitive to multicollinearity
k-Means	Clustering	-	Linear	Unsupervised, sensitive to initial centroids and outliers
PCA	Dimensionality Reduction	-	Linear	Unsupervised, maximizes variance, does not consider class labels
Hierarchical Clustering	Clustering	-	Non-linear	Unsupervised, builds nested clusters, linkage methods affect results
Perceptron	Classification	Discriminative	Linear	Simple model, only for linearly separable data
Multi-Layer Perceptron	Classification/Regression	Discriminative	Non-linear	Can solve XOR problem, uses backpropagation for training

Week 1

Over the past two decades, several major trends in research and computing have collectively enabled the rise of cloud computing:

1. Virtualization has allowed multiple virtual machines to run efficiently on a single physical host, maximizing hardware utilization—crucial for scalable cloud infrastructure.
2. Distributed computing models, such as grid and cluster computing, demonstrated how large tasks could be executed across multiple machines, forming the architectural basis of the cloud.
3. Broadband Internet access made it feasible to access remote resources with low latency, enabling the core idea of "computing over the internet."
4. Utility computing introduced the pay-as-you-go model, shifting IT investment from capital to operational expenditure, thus attracting cost-conscious enterprises.
5. Scalability and elasticity gave organizations the ability to dynamically adjust resources based on real-time demand, improving flexibility and cost-efficiency.
6. Service-Oriented Architecture (SOA) encouraged modular, reusable components, shaping how cloud services are designed and integrated.
7. Data storage innovations, such as SSDs and software-defined storage, improved data throughput and reliability, meeting the demands of large-scale cloud workloads.
8. Security and privacy advancements, including encryption, access control, and compliance standards, addressed growing concerns around trusting third-party providers.
9. Open source software, like OpenStack and Kubernetes, fueled rapid innovation and made cloud technologies more accessible to startups and researchers alike.
10. Economic pressures to cut costs and improve agility further accelerated cloud adoption, as organizations sought to benefit from economies of scale and specialized expertise.

Week 1

Imperative languages are based on commands, in the form of instructions and statements.

• Commands are executed.
• Commands have an effect, such as to update the computation state,and later code may depend on this update.

Logic programming languages are based on finding values that satisfy a set of constraints.

Week 1

• Machine learning is a method of teaching software to learn from data and make decisions on their own, without being explicitly programmed.

Three ingredients for machine learning

1. Data

• Discrete vs continuous vs ...
• Big data vs small data
• Labeled data vs unlabeled data
• Public vs sensitive data

Week 1

• Performance of a database system comes from Hardware (The speed of the processor, Number of processors, Number of drives and I/O bandwidth, Size of main memory, Communication network, Type of architecture), Software (Type and details of database technology used for a given application). and Database tuning, crash recovery (Indexing parameters, Data duplication, Sharing data)

Week 1: Intro to Projects

Project Characteristics:

• Introduce Change
• Temporary
• Cross-functional
• Deal with Unknown
• Unique
• Vary in Size

First Project Management Process:

• Initialization

PM Skills: