Comparisons and
Roadmaps

Guide to AI and Data Courses
The big tableOften-confused pairsWhich course?From interest to courseRoadmaps
Sercan KÜLCÜ · Giresun University · Computer Engineering

The big table

The Big Table: I. Core

CourseApproachInputOutputSuccess measure
Artificial Intelligencesearch, logic, probability and learning combinedproblem definition: states, actions, goalaction sequence, plan or decisionreaching the goal, cost, rationality
Machine Learningstatistical learninglabeled or unlabeled examplesa model that makes predictionserror on unseen data
Neural Networksnetwork of simple weighted unitsnumeric feature vectorclass or numeric predictionloss and accuracy
Deep Learningmultilayer representation learningraw data: images, text, audiolearned representation and predictionaccuracy, loss, scaling

The big table

The Big Table: II. Branches of AI

CourseApproachInputOutputSuccess measure
NLPlinguistics, statistics and deep learningtextlabel, translation, summary, answeraccuracy, BLEU, human evaluation
Reinforcement Learningtrial and error with rewardsenvironment: state, rewardpolicy: state → actiontotal expected reward
Pattern Recognitionstatistical decision theorymeasurement or feature vectorclass labelclassification error
Fuzzy Logicpartial membership and linguistic rulesmeasurements, expert rulescontrol signal or decisioncontrol quality, expert agreement
Evolutionary Optimizationpopulation-based heuristic searchobjective, solution encodinggood, not proven optimal, solutionquality, evaluation count
Expert Systemssymbolic logic and rulesfacts and expert rulesconclusion with justificationexpert agreement, explainability
Graphical Modelsprobability and graphs combinedvariables, dependencies, dataconditional probability, best explanationlikelihood, inference accuracy
Generative Modelslearning the data distributionhuge, mostly unlabeled datanew samples: images, text, audiorealism and diversity (FID etc.)
Multi-Agent Systemsinteracting autonomous agentsagents, goals, protocolsjoint behavior, equilibriumsocial welfare, stability
Explainable AI & Ethicshuman-centered evaluationmodel, its data and context of useexplanation, audit report, principlesinterpretability, fairness, accountability

The big table

The Big Table: III. Signals, Images and Audio

CourseApproachInputOutputSuccess measure
Signal Processingmathematical transforms and filterssampled signalfiltered signal, spectrumsignal-to-noise ratio, frequency response
Image Processingpixel-level transformationsimageprocessed image or regionsvisual quality, PSNR
Computer Visionextracting meaning from imagesimage or videolabel, box, depth, 3D structureaccuracy, mAP, IoU
Speech and Audioacoustic signal and language modelaudio waveformtext, speaker, sound eventword error rate (WER)
Medical Imagingimage processing plus clinical knowledgeCT, MRI, X-ray, ultrasoundsegmentation, measurement, diagnostic suggestionsensitivity, specificity, Dice
Remote Sensingphysical sensing and image analysismultispectral satellite imageryland cover map, changemap accuracy, kappa
Computer Graphicsgeometry and light simulationscene: models, lights, cameraimage or animationrealism, frame rate
Roboticssense-plan-act loopsensor measurementsmotor commands, pathtask success, safety
Biometricsmatching and verificationface, fingerprint, iris, voiceidentity decisionFAR, FRR, EER

The big table

The Big Table: IV. Data

CourseApproachInputOutputSuccess measure
Data Miningdiscovery-driven analysislarge tables or transaction datarules, clusters, outliersinterestingness, support, confidence
Data Sciencequestion, data, analysis, storyraw, messy datareport, visualization, modelimpact on decisions, accuracy
Big Datadistributed storage and computationhuge, fast-flowing dataprocessed data, query resultsscalability, latency, cost
IR & Recommendersmatching and rankingquery, documents, user historyranked result listprecision, recall, NDCG
Time Seriestime-dependent statistical modelstimestamped measurementsforecast, trend, seasonalityforecast error (MAE, RMSE)
Social Network Analysisgraph theory and statisticsnodes and linkscentrality, communities, diffusionmodularity, interpretability

The big table

The Big Table: V. Mathematical Foundations

CourseApproachInputOutputSuccess measure
Linear Algebravector spaces and transformationsvectors, matricessolution, decomposition, transformationaccuracy, numerical stability
Probability & Statisticsprobability models and inferencesample dataestimate, confidence interval, test resultunbiasedness, power, significance
Optimizationmathematical model of objective and constraintsobjective function and constraintsoptimal or near-optimal solutionobjective value, convergence rate
Information Theorymeasuring information with probabilitysource and channel modelbounds, codes, compression ratebits, channel capacity

The big table

The Big Table: VI. Systems and Applications

CourseApproachInputOutputSuccess measure
MLOpssoftware engineering meets MLtrained model, data pipelinea running, monitored servicelatency, uptime, data drift
Embedded AIresource-constrained inferencetrained model, device constraintssmall, fast modellatency, energy, memory
GPU & Parallel Computingdata-parallel computationparallelizable algorithmaccelerated programspeedup, memory bandwidth
AI in Cybersecuritysecurity knowledge plus MLnetwork traffic, logs, filesalert, class, risk scoredetection rate, false alarm rate
Bioinformaticsbiology, algorithms and statisticsDNA, protein sequences, expression dataalignment, structure, function predictionbiological validity

The big table

Math and Data Needs (1/3)

CourseMath loadData needs
Artificial Intelligencemedium: logic, probability, graphslittle; often a model and knowledge suffice
Machine Learninghigh: linear algebra, probability, optimizationmoderate to large
Neural Networksmedium-high: derivatives, matricesmoderate
Deep Learninghigh: derivatives, matrices, probabilitylarge; needs GPUs
NLPmediumlarge text corpora
Reinforcement Learninghigh: probability, dynamic programminggenerated through interaction
Pattern Recognitionmedium-high: probability, linear algebralabeled examples
Fuzzy Logiclow-medium: set theorylittle; expert knowledge
Evolutionary Optimizationlow-mediumnone; evaluations suffice
Expert Systemslow: propositional and predicate logiclittle; a knowledge base
Graphical Modelshigh: probabilitymoderate
Generative Modelshighvery large
Multi-Agent Systemsmedium: game theorylittle; simulation

The big table

Math and Data Needs (2/3)

CourseMath loadData needs
Explainable AI & Ethicslow-mediumthe model's own data
Signal Processinghigh: complex numbers, Fourierno learning; the signal suffices
Image Processingmedium: convolution, Fourierlittle; learning not required
Computer Visionhigh: geometry, linear algebramany labeled images
Speech and Audiomedium-highmany hours of recordings
Medical Imagingmedium-highlittle; labels are expensive
Remote Sensingmediumplenty; labels limited
Computer Graphicshigh: linear algebra, geometrynone
Roboticshigh: transformations, controlphysical experiments and simulation
Biometricsmediumfew samples per person
Data Miningmediumlarge
Data Sciencemedium: statisticsmoderate to large
Big Datalowvery large

The big table

Math and Data Needs (3/3)

CourseMath loadData needs
IR & Recommendersmedium: vector spaceslots of interaction data
Time Seriesmedium-highlong history
Social Network Analysismedium: graphs, matricesnetwork data
Linear Algebrathe theory itselfnone
Probability & Statisticsthe theory itselfa sample
Optimizationthe theory itselfnone
Information Theoryhigh: probabilitynone
MLOpslowstreaming production data
Embedded AImediumtraining data already exists
GPU & Parallel Computinglow-mediumnone
AI in Cybersecuritymediumimbalanced and constantly changing
Bioinformaticsmedium-highhuge sequence datasets

Pairs

Often-Confused Pairs (1/2)

PairDifference
Image Processing ↔ Computer VisionIn image processing both input and output are images; in computer vision the input is an image and the output is meaning: a label, a box, depth.
Artificial Intelligence ↔ Machine LearningAI can find how to reach a goal by search or inference; ML learns the same behavior by generalizing from data.
Machine Learning ↔ Data MiningThey share algorithms; ML aims to predict new examples, data mining to discover patterns in data.
Data Science ↔ Data MiningData mining focuses on finding patterns in large data; data science covers the whole path from question to decision.
Neural Networks ↔ Deep LearningDeep learning moves on to convolutional, recurrent and transformer architectures; this course builds the basic network and training they rest on.
Pattern Recognition ↔ Machine LearningML studies learning algorithms in general; this course focuses on the design from measurement to class decision and on the Bayes error.
NLP ↔ Speech and AudioSpeech recognition goes from the sound wave to text; NLP analyzes meaning once the text is available.

Pairs

Often-Confused Pairs (2/2)

PairDifference
Generative Models ↔ Deep LearningDeep learning is the general toolkit and mostly makes predictions; this course uses those tools to generate new samples from a data distribution.
Computer Graphics ↔ Computer VisionComputer vision tries to recover the scene from an image; computer graphics produces the image from the scene.
Machine Learning ↔ Probability & StatisticsStatistics emphasizes a model's meaning and uncertainty, ML its predictive performance.
IR & Recommenders ↔ Data MiningData mining discovers patterns in data; this course turns discovery into a ranked list for one user's current need.
Expert Systems ↔ Fuzzy LogicFuzzy logic lets rules be partly true; this course teaches representing knowledge and chaining rules in general.
Evolutionary Optimization ↔ OptimizationOptimization reaches a provable optimum using derivatives and convexity; this course teaches derivative-free search with no guarantee but wide applicability.
MLOps ↔ Big DataBig data systems store and process data at scale; this course manages the lifecycle of models trained on that data.

Which course?

Which Course? (1/2)

For each problem, decide which course it belongs to, then open the answer.

  1. Automatically outlining tumours in a hospital's MRI scans.
    AnswerMedical Imaging, Computer Vision, Image Processing. Segmentation comes from image processing and computer vision; clinical data, scarce labels and accountability make it medical image processing.
  2. Building a "customers who bought this also bought" list for an online shop.
    AnswerIR & Recommenders, Data Mining. Association rules come from data mining; presenting a ranked list to the user is the job of recommender systems.
  3. Controlling an oven's temperature with rules like "a bit hot" and "very hot".
    AnswerFuzzy Logic. Linguistic rules without sharp boundaries, used for control, are the classic application of fuzzy logic.
  4. A drone learning to stay level in the wind by trying.
    AnswerReinforcement Learning, Robotics. No labels, only a reward: reinforcement learning. The physical body and its sensors are the business of robotics.
  5. Storing and querying per-second readings from a hundred thousand sensors.
    AnswerBig Data, Time Series. Streaming data too large for one machine belongs to big data systems; analysing it over time is time series work.

Which course?

Which Course? (2/2)

For each problem, decide which course it belongs to, then open the answer.

  1. Finding the shortest total route plan for school buses.
    AnswerEvolutionary Optimization, Optimization, Artificial Intelligence. A combinatorial optimization problem: exact methods come from optimization, good solutions to large instances from metaheuristics.
  2. Auditing whether a credit model treats some groups of applicants unfairly.
    AnswerExplainable AI & Ethics, Data Science. Fairness measures, explanations and accountability are the subject of explainable AI and ethics.
  3. Voice commands that work on a phone without an internet connection.
    AnswerEmbedded AI, Speech and Audio. The recognition model comes from speech and audio processing; fitting it onto the device is the job of embedded AI.
  4. Finding the genes linked to a disease from expression data of thousands of samples.
    AnswerBioinformatics, Probability & Statistics. Biological data and its interpretation are bioinformatics; significance across many tests needs statistics.
  5. Catching unusual connections in a bank's network traffic.
    AnswerAI in Cybersecurity, Data Mining. Outlier detection comes from data mining; an attacker who adapts makes it a cybersecurity problem.

Choosing

From Interest to Course

Pick one or more interests; the courses that serve them are listed, ranked by how many of your interests they touch.

Choosing

Roadmaps (1/2)

Vision track Systems that make sense of images
Language track Systems that work with text and speech
Data track Turning data into decisions and forecasts

Each track runs from prerequisites to specialization; the tracks cross, and following more than one is normal.

Choosing

Roadmaps (2/2)

Robotics track Sensing and acting in the physical world
Production track Turning a model into a real product

Each track runs from prerequisites to specialization; the tracks cross, and following more than one is normal.

Choosing

Final Thoughts

A common language

Linear algebra, probability and optimization are shared by almost every course; with these in place, moving between courses is easy.

The hub

Machine learning sits at the centre of the map: the branches rest on it and the applications use it.

The question decides

Two courses may use the same algorithm; what separates them is the question they ask and how they measure success.

The field moves

Separate courses today may merge tomorrow; understanding the kinds of relation lasts longer than memorizing the map.

The End

Chapter 40: Comparisons and Roadmaps
← Previous chapter · Map of the field
1 / 1 Sercan KÜLCÜ. All rights reserved.