Computer Abstractions & Technology + Machine Learning Practice Flashcards

A complete set of flashcards covering vocabulary from Computer Architecture performance metrics, Python data science libraries, and Supervised Learning techniques (Regression and Classification).

Personal computers (PCs)

General purpose computers that run a variety of software and are designed around a cost/performance tradeoff.

Server computers

Network-based computers that emphasize high capacity, performance and reliability, ranging from small servers to building-sized units.

Supercomputers

A type of server designed for high-end scientific and engineering calculations; they represent the highest capability but represent a tiny fraction of the market.

Embedded computers

Computers hidden inside other systems that have stringent power, performance, and cost constraints.

Personal Mobile Devices (PMDs)

Battery-operated, internet-connected devices costing hundreds of dollars, such as smartphones, tablets, and e-glasses.

Cloud computing

Computing that runs on Warehouse-Scale Computers (WSC) and delivers Software as a Service (SaaS).

Instruction Set Architecture (ISA)

The hardware/software interface of a computer system.

Application Binary Interface (ABI)

The combination of the Instruction Set Architecture (ISA) and the system-software interface.

Volatile main memory

Memory that loses its contents when power is turned off.

Non-volatile secondary memory

Memory that retains contents without power, such as magnetic disks, flash memory, and optical disks.

Yield

The proportion of working dies per wafer in integrated circuit manufacturing.

Response time

Also known as latency; it is the measure of how long a task takes to complete.

Throughput

The total amount of work done per unit time, such as tasks per hour.

Relative performance formula

$$\text{Performance} = \frac{1}{\text{Execution Time}}$$

CPU time

The time the CPU spends on a specific job, excluding I/O and other jobs' shares, calculated as $\text{user CPU time} + \text{system CPU time}$.

Clock period

The duration of one clock cycle.

Clock rate

Also known as frequency, defined as $\text{Clock rate} = \frac{1}{\text{Clock period}}$. Units are typically in Hz.

CPU Time (Core Formula)

$$\text{CPU Time} = \frac{\text{CPU Clock Cycles}}{\text{Clock Rate}}$$

CPI

Average Cycles Per Instruction; the average number of clock cycles per instruction set by the CPU hardware.

Full Performance Equation

$\text{CPU Time} = \text{IC} \times \text{CPI} \times \text{Clock Cycle Time}$ where IC is the Instruction Count.

Dynamic Power in CMOS

$$\text{Power} = \text{Capacitive load} \times \text{Voltage}^2 \times \text{Frequency}$$

Amdahl's Law

A principle stating that the overall speed-up from improving one part of a system is limited by the unaffected part: $T_{\text{improved}} = \frac{T_{\text{affected}}}{\text{improvement factor}} + T_{\text{unaffected}}$

MIPS

Millions of Instructions Per Second; a performance metric calculated as $\text{MIPS} = \frac{\text{Instruction Count}}{\text{Execution Time} \times 10^6}$.

NumPy

A Python library used for numerical arrays and descriptive statistical analysis such as mean, median, and standard deviation.

Pandas

A Python library used for loading external data and managing data tables (DataFrames).

Supervised learning

Learning from data where correct labels or outcome values are known, used to train a model to predict future outcomes.

Regression

A supervised learning task where the outcome variable is continuous.

Line of Best Fit

The linear regression formula $y = c + m \times X$ where $c$ is the constant/intercept and $m$ is the slope/coefficient.

Mean Squared Error (MSE)

A measure of fit for regression calculated as $\text{MSE} = \frac{1}{n} \times \text{sum}_{i=1}^{n} (y_i - \text{yhat}_i)^2$.

RMSE

Root Mean Squared Error, calculated as $\text{RMSE} = \text{sqrt}(\text{MSE})$; it is expressed in the same units as the outcome variable $y$.

Ridge Regression

A type of regularization using an L2 penalty: $\text{Ridge Loss}(y, \text{yhat}) = \text{MSE}(y, \text{yhat}) + \text{alpha} \times \text{sum}_{i=1}^{m} \text{theta}_i^2$.

Lasso Regression

A type of regularization using an L1 penalty: $\text{Lasso Loss}(y, \text{yhat}) = \text{MSE}(y, \text{yhat}) + \text{alpha} \times \text{sum}_{i=1}^{m} |\text{theta}_i|$; it can perform automatic feature selection.

Gradient descent

A robust, iterative learning procedure that updates parameters by moving opposite to the gradient of the error surface to find the minimum error.

Classification

A supervised learning task used to predict discrete categories or classes.

k-Nearest Neighbours (k-NN)

A classification algorithm that assigns a label to a new data point based on the majority vote of the $k$ most similar existing data points.

Decision Trees

A classification or regression model built as a tree structure consisting of decision nodes (split points) and leaf nodes (class labels).

Entropy

A measure of a sample's homogeneity; $E(S) = - \text{sum}_i p_i \times \text{log}_2 p_i$.

Information Gain

The difference between pre-split entropy and post-split entropy: $\text{Gain} = E(\text{before split}) - E(\text{after split})$.

Random Forest

An ensemble of many decision trees built using bootstrap samples and random variable selection at each node to reduce overfitting.

Sigmoid function

A function used in logistic regression to squash real values into a probability range between 0 and 1: $g(z) = \frac{1}{1 + e^{-z}}$.

Softmax function

A function used in multinomial regression to normalize a vector of $n$ real numbers into a distribution of $n$ probabilities that sum to 1.

Naïve Bayes

A classifier based on Bayes' theorem that assumes independence among predictors.

Bayes' Theorem

The formula for computing posterior probability: $P(c | x) = \frac{P(x | c) \times P(c)}{P(x)}$.

Support Vector Machine (SVM)

A classifier that maps data to a higher dimension to find a linear optimal separating hyperplane.

Support vectors

The data points closest to the separating hyperplane in a Support Vector Machine model.

Margin

The distance between the separating hyperplane and the support vectors, which SVM aims to maximize.