antipatternCanonicalName,sources,labels,subcategory,category
dead experimental codepaths,"P01, P05, P07",,ML-specific,code antipatterns
ML code 'dead end',P10,,ML-specific,code antipatterns
complex container comprehension ,P08,,Recurrent in ML-systems,code antipatterns
long lambda function,P08,,Recurrent in ML-systems,code antipatterns
long ternary conditional expression,P08,,Recurrent in ML-systems,code antipatterns
Constant data values,P21,anomalies,data anomalies,data antipatterns
Infinity data values,P21,anomalies,data anomalies,data antipatterns
Uncommon sign,P15,anomalies,data anomalies,data antipatterns
Unnormalized feature,"P15, P18",outliers,data anomalies,data antipatterns
Extreme outliers,P15,outliers,data anomalies,data antipatterns
List length outliers,P15,outliers,data anomalies,data antipatterns
underutilized data dependencies,"P01, P19",dependencies,data dependencies,data antipatterns
unstable data dependencies,"P01, P03, P19","dependencies, evolution",data dependencies,data antipatterns
data silos,P13,infrastructure,data management,data antipatterns
no privacy controls in pipeline,P02,infrastructure,data management,data antipatterns
unsound data storage,P07,infrastructure,data management,data antipatterns
premature data aggregation,P17,management,data management,data antipatterns
unmanaged data heterogeneity,P20,management,data management,data antipatterns
undocumented data collection process,P20,"management, documentation",data management,data antipatterns
unsound/missing metadata,"P14, P17, P20","metadata, documentation",data management,data antipatterns
inaccurate data,"P04, P17",data quality,data quality,data antipatterns
Data duplication,"P15, P17, P20, P21",duplication,data quality,data antipatterns
Data miscoding smell,"P01, P15, P17, P18",miscoding,data quality,data antipatterns
Null/missing data values,"P15, P17, P20, P21",missing entries,data quality,data antipatterns
emphasis on available data,P04,relevance,data relevance,data antipatterns
overcurated data,P10,relevance,data relevance,data antipatterns
stale data,P10,"relevance, evolution",data relevance,data antipatterns
glue code,"P01, P05, P12",integration,integration,design & architecture antipatterns
pipeline jungle,"P01, P05",integration,integration,design & architecture antipatterns
undeclared consumers,"P01, P16, P19",modularity,modularity,design & architecture antipatterns
monolithic ML pipeline,P13,modularity,modularity,design & architecture antipatterns
prototype as architectural foundation,P01,"modularity, evolution",modularity,design & architecture antipatterns
correction cascades,"P01, P04, P19",modularity,modularity,design & architecture antipatterns
multiple-language smell,"P01, P05, P06",technology,technology adoption,design & architecture antipatterns
ad-hoc solution,P13,technology,technology adoption,design & architecture antipatterns
Undefined human accountability,P04,accountability,,ethics antipatterns
biased models,"P02, P05",biased results,,ethics antipatterns
unmanaged social impact,P04,social impact,,ethics antipatterns
black-box models ,P11,transparency,,ethics antipatterns
inadequate configuration management,P07,configuration,,infrastructure antipatterns
Weak or missing debugging,P09,"debugging, production",,infrastructure antipatterns
missing deployment mechanism,P16,deployment,,infrastructure antipatterns
Weak or missing monitoring,"P02, P09, P12","monitoring, production",,infrastructure antipatterns
feature entanglement,"P01, P05, P16, P19",features,features,model antipatterns
epsilon features ,"P02, P05",features,features,model antipatterns
biased feature selection,P10,features,features,model antipatterns
bundled features,P05,features,features,model antipatterns
features violating meta-level requirements,P02,features,features,model antipatterns
too many features,P02,features,features,model antipatterns
no code review for models ,P02,management,management,model antipatterns
no version control for models ,P02,management,management,model antipatterns
unrecognized adversarial agents,P04,management,management,model antipatterns
Backwards-incompatible model,P06,"management, evolution",management,model antipatterns
spurious correlation,"P04, P10",training,training,model antipatterns
data leakage,P17,training,training,model antipatterns
homogeneous training data,P17,training,training,model antipatterns
low model quality on specific data slices,P02,training,training,model antipatterns
numerically unstable model,P02,training,training,model antipatterns
training not reproducible,P02,training,training,model antipatterns
untuned hyperparameters,P02,training,training,model antipatterns
Weak missing value handling,P18,training,training,model antipatterns
direct feedback loops,"P01, P16, P17","training, evolution",training,model antipatterns
data drift,"P16, P17",training/serving skew,training/serving skew,model antipatterns
distribution skew,P14,training/serving skew,training/serving skew,model antipatterns
feature skew,P14,training/serving skew,training/serving skew,model antipatterns
scoring/serving skew,P14,training/serving skew,training/serving skew,model antipatterns
stale models,"P02, P04","training/serving skew, evolution",training/serving skew,model antipatterns
training/serving skew,"P02, P07, P10, P12, P16, P17","training/serving skew, production",training/serving skew,model antipatterns
potential for model misuse,P11,validation,validation,model antipatterns
offline/online proxy metric divergence,"P02, P16","validation, production",validation,model antipatterns
fixed thresholds in dynamic systems,P01,"validation, production",validation,model antipatterns
proxy goals,P10,"validation, production",validation,model antipatterns