Code and data, from various projects:

• (2026) Flatland barriers code (under construction) (when does memory/learning help a navigating agent? see the Flatland page )
• (2026) Python code for sparse moving averages (SMAs), for the TMLR paper, "Lifelong Open-Ended Probability Predictors" (and earlier "Tracking Changing Probabilities.." ).
• A dataset of 22 graphs (our work at Cisco), is now available at the Stanford SNAP (thanks to Rok Sosic), also available at UCI ML repository. It contains edges (TCP/UDP) from distributed applications. Two graphs have reference groupings of nodes (ground truth). This IWSPA 2022 paper describes the data (note: another graph was added after paper publication) (the README file, and basic Python code made available to read the data).
• A dataset of multimodal feature vectors, YouTube Multiview Video Games, available at UCI ML repository, specially useful for multiview (multimodal) machine learning research (also here in smaller partitions), following our MLJ 2013 work.
• A Python implementation of sparse multiclass EMA, a version of (sparse) index learning, suitable for non-stationary many-class problems (thanks to Jose Antonio), following our IJCAI 2009 paper (note: the above SMAs code is a recent development focusing on probability prediction and goes beyond sparse EMA, while this version is for multiclass ranking/accuracy objectives).