Supported Algorithms

librec-auto supports all of the algorithms in Librec. As of September 15, 2020, the following algorithms are supported:

aobpr
aspectmodelranking
aspectmodelrating
associationrule
asvdpp
autorec
bhfree
biasedmf
biasedmf
bipolarslopeone
bnppf
bpmf
bpoissmf
bpr
bucm
cdae
climf
cofiset
constantguess
convmf
cptf
dlambdafm
eals
efm
ensemble-linear
ensemble-stacking
external
ffm
fismauc
fismrmse
fmals
fmftrl
fmsgd
gbpr
globalaverage
gplsa
hft
hybrid
irrg
itemaverage
itembigram
itemcluster
itemknn
itemknn
lda
ldcc
librec-default
librec
listrankmf
llorma
mfals
mostpopular
nmf
nmfitemitem
personalitydiagnosis
pitf
plsa
pmf
pnmf
prankd
randomguess
rankals
rankgeofm
rankpmf
ranksgd
rbm
remf
rfrec
rste
sbpr
slim
slopeone
socialmf
sorec
soreg
svdpp
tfidf
timesvd
topicmfat
topicmfmt
trustmf
trustsvd
urp
useraverage
usercluster
userknn
usg-test
wbpr
wrmf

Custom algorithms

In future releases, you will be able to add your own algorithms to librec-auto.

Supported metrics

librec-auto supports all of the metrics in Librec. As of September 15, 2020, the following metrics are supported:

auc
ap (Average Precision)
arhr (Average Reciprocal Hit Rate)
diversity
hitrate
ndcg
precision
recall
rr (Reciprocal Rank)
featurediversity
novelty
entropy
giniindex
icov (Item Coverage)
mae
mpe
mse
rmse
csp (Consumer-side Statistical Parity of nDCG)
psp (Provider-side Statistical Parity of exposure)
miscalib (Miscalibration)
dppf (Discounted Proportional Provider-side Fairness)
dpcf (Discounted Proportional Consumer-side Fairness)
nonpar (NonParityUnfairness)
valunfairness (Value Unfairness)
absunfairness (Absolute Unfairness)
overestimate (Overestimation Unfairness)
underestimate (Underestimation Unfairness)
ppr (PPercent Rule as applied to provider exposure)

Custom metrics

To add a new metric, see discussion in python-side