docs/CHANGELOG.md

In ajwills72/catlearn: Formal Psychological Models of Categorization and Learning

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog

Version 1.0 [cran]

Added

• slpDGCM [model] introduce the implementation of similarity-dissimilarity general-context model (Stewart and Morin, 2007), that also includes settings to run the version of O'Bryan et al. (2018). (@lenarddome)

Changed

• slpSUSTAIN [model] Enhancement: add probability of item being reported as 'old'.

Version 0.9.1

Added

• slpNNCAG, slpNNRAS [model] Added Models 4 and 5 from Paskewitz & Jones (2020). (@lenarddome)

Changed

• slpEXIT [code] 4af799d23adb8f8d42e2479adcb64855709e24bc Minor tweak to fix linux+clang14 warning (@ajwills72)

Version 0.8.3

Changed

• DESCRIPTION Introduced package version requirement for RcppArmadillo to avoid cryptic install errors where installing on out-of-date systems. (@ajwills72)

Version 0.8.2

Added

• slpLMSnet [model] Added ratio-rule response output option in response to user query (Ivan Ivanchei) (@lenarddome)

Changed

• slpLMSnet [model] Documentation improved in response to user query (Ivan Ivanchei) (@lenarddome)

Version 0.8.1

Added

• slpNNCAG [model] and slpNNRAS [model] Added Paskewitz and Jones (2020) Models 4 and 5, which are fragmented versions of Kruschke's EXIT model (@lenarddome)

Version 0.7.5

Added

• slpLMSnet [model] Added Gluck and Bower (1988) adaptive LMS-network extension of the Rescorla & Wagner (1972) rule (aslo called LMS or delta learning rule)(@lenarddome)

Version 0.7.4

Changed

• slpSUSTAIN [fixed] a bug relating to new default behaviour ties = "random", introduced in Version 0.7.2. The bug was that a random cluster was selected even in the absence of ties. This patch removes the error, with the code now working as documented (@lenarddome)

Version 0.7.3

Added

• slpALCOVE [error checking] - Throw error if (1) alpha and h don't match in size, or (2) h and w don't match in size, or (3) c, phi, la, lw are outside allowed values, or (4) width of tr does not match model spec in st. Warn if: (5) phi exceeds 20, (6) r or q have a value other than 1 or 2. @ajwills72

Version 0.7.2

Added

• slpSUSTAIN [error checking] - colskip: if it is too large, the user will get a more informative error (@lenarddome)
• slpSUSTAIN [error checking] - dimensions: if number of dimensions don't match up (lambda, w, cluster, or dims), receive an error message (@lenarddome)
• slpSUSTAIN [new features] - tied activations: the model can select the cluster that was recruited earlier, or the model can select clusters with tied activations recently. See the documentation for slpSUSTAIN. (@lenarddome)

Changed

• slpSUSTAIN [fix] - cluster recruitment: on rare cases, the model might make the wrong response, and tries to recruit a cluster even though the cluster already exist. This has been fixed, see documentation notes. (@lenarddome)
• slpSUSTAIN [fix] - tied activations: on some cases, more than one cluster can have the same activations. The new default behaviour for slpSUSTAIN is to select one cluster randomly from those that have the same activations, but it can be changed by setting ties = "first" in its arguments. See documentation. (@lenarddome)
• slpSUSTAIN [code] - code chunks: included code chunks for easier inspection

Removed

• NONE

ajwills72/catlearn documentation built on Feb. 10, 2024, 4:28 a.m.