Open-source software
Libraries and research code from my projects, free to use and build on.
Latest release
mcdakit
Multi-criteria decision analysis for Python: 17 ranking methods including TOPSIS, VIKOR, PROMETHEE, ELECTRE, AHP and SPOTIS, rank-reversal-free ranking, and weight sensitivity analysis solved to a threshold. One dependency (numpy).
from mcdakit import Criterion, rank, sensitivity criteria = [ Criterion("Price", weight=0.40, direction="cost"), Criterion("Quality", weight=0.35, direction="benefit"), Criterion("Lead time", weight=0.25, direction="cost"), ] result = rank(matrix, criteria, method="spotis") report = sensitivity(result) # would the ranking survive other weights?
Federated learning and data mining
Frameworks and research code from my projects.
Decentrai
Open-source framework for hierarchical federated learning: clients train locally, regional coordinators aggregate their updates, and a global orchestrator builds the next model. Includes 18 aggregation algorithms, differential privacy and LoRA fine-tuning of language models.
FEDn
Open-source, framework-agnostic federated learning system for cross-silo and cross-device training. Co-developed at Uppsala University with Scaleout Systems.
FedQAS
Privacy-aware machine reading comprehension with federated learning, a FEDn client for training question-answering models across sites.