AWS Certified AI Practitioner(26) - Model Fit, Bias, and Variance
π Model Fit, Bias, and VarianceWhen a machine learning model performs poorly, one of the first thingsto check is whether itβs a good fit for the data. This is oftendiscussed in terms of overfitting, underfitting, andbalance. β
Model FitπΉ Overfitting The model performs very well on training data.\ But performs poorly on evaluation or unseen test data.\ Example: A line that connects every single training point perfectlyβ great for training, useless for new data.\ Common when the model is too...
AWS Certified AI Practitioner(25) - Reinforcement Learning
π§ Reinforcement Learning (RL) & RLHF1. What is Reinforcement Learning (RL)?Reinforcement Learning is a type of machine learning where an agentlearns to make decisions by interacting with an environment andmaximizing rewards. Agent β the learner or decision-maker (e.g., a robot, softwarebot).\ Environment β the external system the agent interacts with(e.g., a maze, stock market).\ State β the current situation of the environment.\ Action β the choice the agent makes.\ Reward β feedback (...
(νκ΅μ΄) AWS Certified AI Practitioner (25) - κ°ννμ΅
π§ κ°ννμ΅(Reinforcement Learning, RL)κ³Ό RLHF μ½κ² μ΄ν΄νκΈ°1. κ°ννμ΅μ΄λ?κ°ννμ΅(RL)μ νκ²½(Environment) μμμ **μμ΄μ νΈ(Agent)**κ°νλ(Action)μ μννλ©΄μ 보μ(Reward)μ μ»κ³ , μ₯κΈ°μ μΌλ‘ λμ 보μμκ·Ήλννλ λ°©μμΌλ‘ νμ΅νλ λ¨Έμ λ¬λ κΈ°λ²μ
λλ€. ν΅μ¬ κ°λ
Agent: νμ΅μ λλ μμ¬κ²°μ μ (μ: λ‘λ΄) Environment: μμ΄μ νΈκ° μνΈμμ©νλ μΈλΆ μμ€ν
(μ: λ―Έλ‘) Action: μμ΄μ νΈκ° μ νν μ μλ νλ (μ: μ, μλ, μΌμͺ½,μ€λ₯Έμͺ½ μ΄λ) Reward: νλμ κ²°κ³Όμ λ°λ₯Έ νΌλλ°± (μ: +100 μ = μ±κ³΅, -10μ = λ²½ μΆ©λ) State: νκ²½μ νμ¬ μν (μ: λ‘λ΄μ μμΉ)\ Policy: μνμ λ°λΌ μ΄λ€ νλμ ν μ§ μ νλ μ λ΅ π μν ν¬μΈνΈ: RLμ 보μ(Reward) κΈ°λ° νμ΅μ΄λΌλ μ μ΄ μ€μν©λλ€. 2. κ°ννμ΅ λμ λ°©μ ...
(νκ΅μ΄) AWS Certified AI Practitioner (24) - λΉμ§λ νμ΅ & μκΈ° μ§λ νμ΅
π€ λ¨Έμ λ¬λ μκ³ λ¦¬μ¦ β λΉμ§λ νμ΅(Unsupervised Learning) & μκΈ° μ§λ νμ΅(Self-Supervised Learning)1. λΉμ§λ νμ΅ (Unsupervised Learning) λΌλ²¨(μ λ΅)μ΄ μλ λ°μ΄ν°μμ ν¨ν΄, ꡬ쑰, κ΄κ³λ₯Ό μ€μ€λ‘ λ°κ²¬νλ νμ΅ λ°©μ λ¨Έμ λ¬λ λͺ¨λΈμ΄ λ°μ΄ν° κ·Έλ£Ήμ λ§λ€κ³ , μ¬λμ΄ κ·Έ κ²°κ³Όμ μλ―Έ(λΌλ²¨)λ₯Ό λΆμ¬ λν κΈ°λ²: ν΄λ¬μ€ν°λ§(Clustering) μ°κ΄ κ·μΉ νμ΅(Association Rule Learning) μ΄μμΉ νμ§(Anomaly Detection) νμ© μ¬λ‘ κ³ κ° μΈλΆν (λ§μΌν
) μΆμ² μμ€ν
μ¬κΈ° νμ§ π μν ν¬μΈνΈ: λΉμ§λ νμ΅μ λΌλ²¨μ΄ μλ λ°μ΄ν°λ₯Ό νμ©νλ€λ μ κΈ°μ΅νκΈ° 2. ν΄λ¬μ€ν°λ§ (Clustering) μ μ: μ μ¬ν νΉμ§μ κ°μ§ λ°μ΄ν°λ₯Ό λ¬Άμ΄ κ·Έλ£Ή(cluster)μΌλ‘ λλ μ¬λ‘: κ³ κ° μΈλΆν(Customer Segmentation) μν©: e-commer...
AWS Certified AI Practitioner(24) - Unsupervised & Semi/Self-Supervised Learning
π§ Machine Learning Algorithms β Unsupervised & Semi/Self-Supervised Learning1. What is Unsupervised Learning? Definition: Machine learning on unlabeled data (no predefined outputs). Goal: Discover hidden patterns, structures, or relationships in the data. Key point: The algorithm finds groups or rules by itself, while humans later assign meaning (labels) to those groups. Common techniques: Clustering β finding groups of similar data (e.g., customer segmentation) Association ...
AWS Certified AI Practitioner(23) - Training Data & Feature Engineering
π Training Data & Feature EngineeringWhy Training Data Matters To build a reliable ML model, you need good quality data.\ Principle: Garbage In β Garbage Out. If your input data is messyor incorrect, your model will produce poor predictions.\ Data preparation is the most critical stage of ML.\ The way you model your data (e.g., labeled/unlabeled,structured/unstructured) directly impacts which algorithms you canuse. π Exam Tip: Expect questions about labeled vs.Β unlabeled ands...
(νκ΅μ΄) AWS Certified AI Practitioner (23) - λ¨Έμ λ¬λ νμ΅ λ°μ΄ν° μ 리
π λ¨Έμ λ¬λ νμ΅ λ°μ΄ν° μ 리1. νμ΅ λ°μ΄ν°(Training Data)μ μ€μμ± μ’μ λ°μ΄ν°λ₯Ό κ°μ ΈμΌ μ’μ λͺ¨λΈμ λ§λ€ μ μμ Garbage In β Garbage Out : μλͺ»λ λ°μ΄ν°λ₯Ό λ£μΌλ©΄ κ²°κ³Όλ μλͺ»λ¨ κ°μ₯ μ€μν λ¨κ³ = λ°μ΄ν°λ₯Ό κΉ¨λνκ² μ€λΉνλ κ² λ°μ΄ν°μ μ’
λ₯μ λ°λΌ μ¬μ©ν μ μλ μκ³ λ¦¬μ¦λ λ¬λΌμ§ 2. λΌλ²¨λ§ λ°μ΄ν° vs λΉλΌλ²¨λ§ λ°μ΄ν°πΉ λΌλ²¨λ§ λ°μ΄ν° (Labeled Data) μ
λ ₯κ°(Input) + μ λ΅(Output Label)μ΄ ν¨κ» μλ λ°μ΄ν° μ: κ³ μμ΄, κ°μμ§ μ΄λ―Έμ§μ κ°κ°μ λΌλ²¨μ΄ ν¨κ» μμ μ¬μ© μ¬λ‘: μ§λνμ΅(Supervised Learning) πΉ λΉλΌλ²¨λ§ λ°μ΄ν° (Unlabeled Data) μ
λ ₯κ°λ§ μκ³ μ λ΅ λΌλ²¨μ΄ μμ μ: κ³ μμ΄/κ°μμ§ μ¬μ§λ§ μκ³ λΌλ²¨μ΄ μλ κ²½μ° μ¬μ© μ¬λ‘: λΉμ§λνμ΅(Unsupervised Learning) β ν¨ν΄μ΄λ κ΅°μ§ μ°ΎκΈ° ...
(νκ΅μ΄) AWS Certified AI Practitioner (22) - 곡μ§λ₯(AI), λ¨Έμ λ¬λ(ML), λ₯λ¬λ(DL), μμ±ν AI (GenAI) μ 리
π€ μΈκ³΅μ§λ₯(AI), λ¨Έμ λ¬λ(ML), λ₯λ¬λ(DL), μμ±ν AI (GenAI) μ 리1. μΈκ³΅μ§λ₯(AI)λ? AIλ μΈκ°μ μ§λ₯μ΄ νμν μΌμ λμ μνν μ μλ μμ€ν
μ λ§λλ κ΄λ²μν κΈ°μ λΆμΌμ
λλ€. μ£Όμ κΈ°λ₯: μΈμ(Perception) μΆλ‘ (Reasoning) νμ΅(Learning) λ¬Έμ ν΄κ²°(Problem Solving) μμ¬κ²°μ (Decision Making) π μν ν¬μΈνΈ: AIλ ν° κ°λ
(μ°μ°)μ΄κ³ , κ·Έ μμ ML β DL β GenAI μμλ‘ μΈλΆ κΈ°μ μ΄ ν¬ν¨λ©λλ€. --- 2. AIμ κ΅¬μ± μμ λ°μ΄ν° κ³μΈ΅: λλμ λ°μ΄ν°λ₯Ό μμ§ ML νλ μμν¬ λ° μκ³ λ¦¬μ¦ κ³μΈ΅: λ°μ΄ν° κ³Όνμμ μμ§λμ΄κ° μꡬμ¬νμ λ§λ μκ³ λ¦¬μ¦ μ€κ³ λͺ¨λΈ κ³μΈ΅: λͺ¨λΈ ꡬ쑰 μ€κ³, νλΌλ―Έν° λ° μ΅μ ν ν¨μ μ μ© β νμ΅ μν μ ν리μΌμ΄μ
κ³μΈ΅: νμ΅λ λͺ¨λΈμ μ€μ μ¬μ©μμκ² μλΉμ€ 3. λ¨Έμ λ¬λ(ML)λ? AIμ ν λΆμΌλ‘...
AWS Certified AI Practitioner(22) - Understanding AI, ML, DL, and GenAI
π€ Understanding AI, ML, DL, and GenAI1. What is Artificial Intelligence (AI)?Artificial Intelligence (AI) is a broad field focused on building intelligent systems capable of tasks that usually require human intelligence, such as: Perception Reasoning Learning Problem solving Decision making π AI is an umbrella term covering multiple techniques. 2. AI Components Data Layer: Collects large amounts of data. ML Framework & Algorithm Layer: Data scientists and engine...
AWS Certified AI Practitioner(21) - PartyRock
π PartyRock β GenAI Playgroundπ What is PartyRock? PartyRock is a Generative AI app-building playground powered by Amazon Bedrock. It allows you to experiment with creating GenAI apps using various Foundation Models (FMs). No coding or AWS account is required. The UI is very similar to Amazon Q Apps, but requires less setup. π Link: PartyRock π Key Features Build GenAI apps without coding Use natural language prompts to create applications Experiment with different widge...