Recenl, cienific dicore in arificial inelligence and daa cience ha foced on eplainable AI (XAI) ih repec o algorihmic ranparenc, inerpreabili, acconabili and finall eplainabili of algorihmic model and deciion. In machine learning, approache can be claified a hie-bo and black-bo. Whie-bo approache, ch a rle learner and indcie programming, rel in eplici model hich are inherenl inerpreable (Rdin, 2019). On he oher hand, black-bo approache, ch a (deep) neral neork, rel in opaqe model. For hi econd pe of model, oer he la ear, differen approache for e-po eplanaion generaion hae been propoed. In hi orkhop, e an o bring ogeher reearch from inerpreable and eplanaor machine learning. Inerpreable ML can profi from recenl propoed eplanaion generaion echniqe o make comple learned model more comprehenible, epeciall o end-er (Mggleon, Schmid e al., 2018; Frnkran, Kliegr, Palheim, 2020; Lonjarre e al., 2020). In pariclar, inerpreable learning can be inegraed ino he conrcion of comple model, e.g., for giding heir conrcion (Ameller e al., 2017), a ell a o refine he repecie model (Weidner, Ameller, Seipel, 2019). Frhermore, i can proide rich rle-baed echniqe o generae inerpreable rrogae model for black-bo learner (Schmid, 2018). Sch rrogae model can be global model generaed b rle-eracion mechanim (Haileilaie, 2016) or local model hich allo richer local eplanaion han imple linear rle a, for inance, propoed b LIME (Rabold e al., 2019). Alo, a fronier direcion i ineigaing pchological phenomena ha can affec he nderanding of machine learning model, ch a cogniie biae and coneraional maim (Kliegr, Bahnik, Frnkran, 2018). Thi inerdiciplinar inpiraion, ch a debiaing echniqe long died b pchologi, ill hopefll conribe o a beer comprehenibili of he rel of model creaed b he ne generaion of machine learning algorihm. XI-ML (Eplainable and Inerpreable Machine Learning) aim a bringing ogeher reearch from inerpreable and eplainable machine learning. Hopefll, inegraing boh area, allo ne perpecie on qeion on appropriae learning formalim, inerpreaion and eplanaion echniqe, heir meric, a ell a he repecie aemen opion arie.
.c lab.cc/ i-ml-2020/) The fir ediion of he XI-ML (Eplainable and Inerpreable Machine Learning) orkhop a held on Sepember 21, 2020 a he 43rd German Conference on Arificial Inelligence, Bamberg, German. The orkhop a deoed o he dicion of he opic menioned aboe. I aimed o proide an inerdiciplinar form o ineigae fndamenal ie in eplainable and inerpreable machine learning a ell a o dic recen adance, rend, and challenge in hee area. From 8 bmiion (6 fll and o hor paper), 5 fll paper and he o hor paper ere acceped for preenaion in a comprehenie reie proce.
The remaining par of he olme preen reied erion of paper ha ere diced dring he orkhop. Borm e al. dic in heir paper abo ho o eplain mliariae ime erie forecaing, in an applicaion o predicing he Sedih GDP. Ne, Mcha e al. preen a poiion paper on ho o conrc paricipaor deign pace for he cone of eplainable AI inerface in eper domain. Afer ha, Volker diced ho he applicaion of he TED (Teaching Eplanaion for Deciion) eplainable AI frameork and he impac of cla (im-)balance. Flei, Bck and Thalmann preen a hor paper on empirical rel in he cone of recriing - abo eplainabili and he inenion o e AI-baed coneraional agen. Poka addree fondaional ie oard oling claificaion problem ih qaniaie abrac argmenaion. Mollenhaer and Ameller preen an approach for eqenial ecepional paern dicoer ing paern-groh (SEPP) - a he bai of an eenible frameork for inerpreable machine learning on eqenial daa. Sn, Chakrabori and Noble dic rel of a comparaie d of eplainer modle in he cone of aomaed kin leion claificaion.
Finall, Marcin P. Joachimiak (Enironmenal Genomic and Sem Biolog Diiion, Larence Berkele Laboraor) kindl agreed o preen a kenoe eniled Ho o each a comper o learn abo microbe ih KG-COVID-19. Thi alk inrodced a ne reorce ha amalgamae SARS-CoV-2 relaed biological knoledge from mliple pecialied knoledge graph and onologie. Wih oer 10 million node, i i one of he large (if no he large) reorce of hi kind. In hi alk, Dr. Joachimiak demonraed he ili of hi reorce for machine learning, emphaiing he need for eplainable echniqe. A m eller, M., Ha a , N., Schmid , A., & Kl pper, B. (2017). E plana ion-a are fea re elec ion ing mbolic ime erie ab rac ion: approache and e perience in a pe ro-chemical prod c ion con e . In IEEE I a a C c I d a I a c (INDIN) (pp. 799-804). IEEE, Bo on, MA, USA F rnkran , J., Kliegr, T., & Pa lheim, H. (2020). On cogni i e preference and he pla ibili of r le-ba ed model . Mac L a , 109 (4), 853-898.
Haile ila ie, T. (2016). R le e rac ion algori hm for deep ne ral ne ork : A re ie . a X a X :1610.05267 .
Kliegr, Tom , p n Bahn k, and Johanne F rnkran . "A re ie of po ible effec of cogni i e bia e on in erpre a ion of r le-ba ed machine learning model ." arXi preprin arXi :1804.02969 (2018).
Lonjarre , C., Robarde , C., Plan e i , M., A b r in, R., & A m eller, M. (2020). Wh Sho ld I Tr Thi I em? E plaining he Recommenda ion of an Model. In IEEE I a a C c Da a Sc c a d A a c . IEEE, Bo on, MA, USA M ggle on, S. H., Schmid, U., Zeller, C., Tamaddoni-Ne had, A., & Be old, T. (2018). Ul ra-S rong Machine Learning: comprehen ibili of program learned i h ILP. Mac L a , 107 (7), 1119-1140. i h erbal c R din, C. (2019). S op e plaining black bo machine learning model for high ake deci ion and e in erpre able model in ead. Na Mac I c , 1 (5), 206-215. Schmid, U. (2018). Ind c i e Programming a Approach o Comprehen ible Machine Learning. In DKB/KIK@ KI (pp. 4-12).
Weidner, D., A m eller, M., & Seipel, D. (2019). Finding Ma imal Non-red ndan A ocia ion R le in Tenni Da a. In D c a a P a a d K d Ma a (pp. 59-78). Springer, Cham.
Edi or ● ● ●
Mar in A m eller, O nabr ck Uni er i , German Tom Kliegr, Uni er i of Economic Prag e, C ech Rep blic
U e Schmid, Uni er i of Bamberg, German Program Commi ee of XI-ML 2020 ● ● ● ● ● ● ● ● ● ● ● ● ● ●