25 lines
2.7 KiB
TeX
25 lines
2.7 KiB
TeX
\begin{WideTab}[t]
|
||
\centering
|
||
\caption{Comparison of OPT with existing standards, policy frameworks, and textbook pillars.}
|
||
\renewcommand{\arraystretch}{1.12}
|
||
\begin{tabular}{@{}p{2.9cm}p{3.1cm}p{3.2cm}p{2.6cm}p{3.0cm}@{}}
|
||
\toprule
|
||
\textbf{Framework / Source} & \textbf{Primary Scope} & \textbf{Unit of Classification} & \textbf{Technique Coverage} & \textbf{Hybrid Handling / Intended Use} \\
|
||
\midrule
|
||
\textbf{OPT (this work)} & Implementation taxonomy & \textit{Operative mechanism} (\Lrn,\ \Evo,\ \Sym,\ \Prb,\ \Sch,\ \Ctl,\ \Swm) with composition grammar & Cross-paradigm (learning, symbolic, probabilistic, search, control, swarm, evolutionary) & Explicit hybrids via \hyb{+}, \hyb{/}, \hyb{\{\,\}}, \hyb{[\,\rightarrow\,]}; designed to interface with risk/process frameworks \\
|
||
\addlinespace[3pt]
|
||
ISO/IEC 22989:2022 \citep{ISO22989} & Concepts \& terminology & Vocabulary / definitions & Technique-agnostic & No hybrid grammar; supports common language across stakeholders \\
|
||
ISO/IEC 23053:2022 \citep{ISO23053} & ML system architecture & Functional blocks (data, training, inference, monitoring) & ML-centric; excludes non-ML pillars (e.g., \Sym,\ \Ctl,\ \Swm) & No explicit hybrid mechanism model; system design/process lens \\
|
||
NIST AI RMF 1.0 \citep{NISTRMF} & Risk management & Risk functions (Govern, Map, Measure, Manage) & Technique-agnostic & No mechanism taxonomy; governance and assurance guidance \\
|
||
NIST AI 200-1 \citep{NISTAI2001} & Use taxonomy & Human–AI task activities & Technique-agnostic & No hybrids; categorizes use contexts for evaluation \\
|
||
OECD AI Classification \citep{OECDClass} & Policy characterization & Multi-axis profile (context, data, model, task) & Broad; includes an “AI model” axis but not a formal mechanism taxonomy & No hybrid grammar; policy comparison and statistics \\
|
||
EU AI Act \citep{EUAIAct} & Regulation (risk-based) & Risk class (prohibited/high/limited/minimal) & Technique-agnostic & Hybrids irrelevant; compliance and obligations \\
|
||
AIMA (Russell \& Norvig) \citep{AIMA4} & Textbook organization & Pillars (search/planning, logic, probabilistic reasoning, learning, agents) & Broad coverage; closest to mechanism families & No standard naming or hybrid code; educational structure \\
|
||
\bottomrule
|
||
\end{tabular}
|
||
|
||
\vspace{4pt}
|
||
\footnotesize \textit{Notes.} OPT supplies a compact, biologically grounded \emph{implementation} taxonomy with a formal hybrid composition code. Standards and policy frameworks remain essential and complementary for vocabulary, lifecycle, risk, management, and regulatory obligations, but they are technique-agnostic or ML-specific and do not provide a mechanism-level naming scheme.
|
||
\label{tab:opt_vs_frameworks}
|
||
\end{WideTab}
|