/
roadmap.html
839 lines (801 loc) · 28.9 KB
/
roadmap.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
<!DOCTYPE html>
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta charset="utf-8" /><meta name="generator" content="Docutils 0.17.1: http://docutils.sourceforge.net/" />
<title>Roadmap — FAT Forensics 0.1.2 documentation</title>
<!-- html title is before nature.css - we use this hack to load bootstrap first -->
<!-- Required meta tags -->
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1, shrink-to-fit=no">
<!-- Bootstrap CSS -->
<link rel="stylesheet" href="_static/css/bootstrap.min.css">
<link rel="stylesheet" href="_static/nature.css" type="text/css" />
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<link rel="stylesheet" type="text/css" href="_static/gallery.css" />
<script type="text/javascript" id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></script>
<script type="text/javascript" src="_static/jquery.js"></script>
<script type="text/javascript" src="_static/underscore.js"></script>
<script type="text/javascript" src="_static/doctools.js"></script>
<script type="text/javascript" src="_static/language_data.js"></script>
<script async="async" type="text/javascript" src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS_SVG"></script>
<link rel="canonical" href="https://fat-forensics.org/roadmap.html" />
<link rel="shortcut icon" href="_static/fatf.ico"/>
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="User Guide" href="user_guide/index.html" />
<link rel="prev" title="Changelog" href="changelog.html" />
<!-- jQuery first, then Bootstrap JS -->
<!-- jQuery is distributed with Sphinx already -->
<!-- <script src="_static/jquery.min.js"></script> -->
<script src="_static/js/bootstrap.min.js"></script>
</head><body>
<div class="container">
<nav class="navbar navbar-expand-md navbar-light bg-light">
<a class="navbar-brand align-text-middle" href="index.html">
<img src="_static/fatf.png"
alt="Logo"
width="35" height="35"
class="d-inline-block align-middle">
FAT Forensics
</a>
<button class="navbar-toggler"
type="button"
data-toggle="collapse"
data-target="#navbarCollapsedContent">
<span class="navbar-toggler-icon"></span>
</button>
<div class="collapse navbar-collapse"
id="navbarCollapsedContent">
<ul class="navbar-nav mr-auto">
<li class="nav-item">
<a class="nav-link" href="index.html">Home</a>
</li>
<li class="nav-item dropdown">
<a class="nav-link dropdown-toggle" href="#" role="button" data-toggle="dropdown">
Documentation
</a>
<div class="dropdown-menu">
<div class="dropdown-header">FAT Forensics</div>
<div class="dropdown-divider"></div>
<a class="dropdown-item" href="getting_started/index.html">Getting Started</a>
<a class="dropdown-item" href="tutorials/index.html">Tutorials</a>
<a class="dropdown-item" href="sphinx_gallery_auto/index.html">Examples</a>
<a class="dropdown-item" href="api.html">API Reference</a>
<a class="dropdown-item" href="how_to/index.html">How-To Guides</a>
<div class="dropdown-divider"></div>
<a class="dropdown-item" href="news.html">News</a>
<a class="dropdown-item" href="development.html">Developers Guide</a>
<a class="dropdown-item" href="contributors.html">Contributors</a>
<a class="dropdown-item" href="changelog.html">Changelog</a>
<a class="dropdown-item" href="#">Roadmap</a>
</div>
</li>
<li class="nav-item">
<a class="nav-link" href="user_guide/index.html">
FAT User Guide
</a>
</li>
</ul>
<div class="search_form form-inline">
<div class="gcse-search" id="cse" style="width: 100%;"></div>
</div>
</div>
</nav>
</div>
<!-- GitHub "fork me" ribbon -->
<!--
<a href="https://github.com/fat-forensics/fat-forensics">
<img class="fork-me"
style="position: absolute; top: 0; right: 0; border: 0;"
src="_static/img/forkme.png"
alt="Fork me on GitHub" />
</a>
-->
<span id="forkongithub">
<a href="https://github.com/fat-forensics/fat-forensics">
Fork me on GitHub
</a>
</span>
<div class="container">
<div class="row">
<div class="col-sm-2">
<div class="container sidebar">
<!--
<div class="sphinxsidebar" role="navigation" aria-label="main navigation">
<div class="sphinxsidebarwrapper">
<p class="logo"><a href="index.html">
<img class="logo" src="_static/fatf.png" alt="Logo"/>
</a></p>
<h3><a href="index.html">Table of Contents</a></h3>
<ul>
<li><a class="reference internal" href="#">Roadmap</a><ul>
<li><a class="reference internal" href="#milestone-1">Milestone 1 ✔</a></li>
<li><a class="reference internal" href="#milestone-2">Milestone 2</a></li>
<li><a class="reference internal" href="#milestone-3">Milestone 3</a></li>
<li><a class="reference internal" href="#milestone-4">Milestone 4</a></li>
</ul>
</li>
</ul>
<h4>Previous topic</h4>
<p class="topless"><a href="changelog.html"
title="previous chapter">Changelog</a></p>
<h4>Next topic</h4>
<p class="topless"><a href="user_guide/index.html"
title="next chapter">User Guide</a></p>
<div role="note" aria-label="source link">
<h3>This Page</h3>
<ul class="this-page-menu">
<li><a href="_sources/roadmap.rst.txt"
rel="nofollow">Show Source</a></li>
</ul>
</div>
<div id="searchbox" style="display: none" role="search">
<h3>Quick search</h3>
<div class="searchformwrapper">
<form class="search" action="search.html" method="get">
<input type="text" name="q" />
<input type="submit" value="Go" />
</form>
</div>
</div>
<script type="text/javascript">$('#searchbox').show(0);</script>
</div>
</div>-->
<!-- Add a link to the 'up' page -->
<!-- Add a links to the 'relative' pages -->
<div class="row rel">
<div class="col-6 rellink
pad-r-1
">
<a class="btn btn-primary btn-sm"
href="changelog.html"
accesskey="P"
role="button">
Previous
<br/>
<span class="smallrellink">
Changelog
</span>
<span class="hiddenrellink
left-button
"
data-container="body">
Changelog
</span>
</a>
</div>
<div class="col-6 rellink
pad-l-1
">
<a class="btn btn-primary btn-sm"
href="user_guide/index.html"
accesskey="N"
role="button">
Next
<br/>
<span class="smallrellink">
User Guide
</span>
<span class="hiddenrellink
right-button
"
data-container="body">
User Guide
</span>
</a>
</div>
</div>
<!-- Add a citation banner -->
<div class="alert alert-info" role="alert" style="font-size: 89%; margin-top: 16px;">
Please <a href="getting_started/cite.html"><b>cite us</b></a> if you use the software.
</div>
<!-- Add a page map -->
<div class="row toc">
<ul>
<li><a class="reference internal" href="#">Roadmap</a><ul>
<li><a class="reference internal" href="#milestone-1">Milestone 1 ✔</a></li>
<li><a class="reference internal" href="#milestone-2">Milestone 2</a></li>
<li><a class="reference internal" href="#milestone-3">Milestone 3</a></li>
<li><a class="reference internal" href="#milestone-4">Milestone 4</a></li>
</ul>
</li>
</ul>
</div>
</div>
</div>
<div class="col col-sm-10">
<div class="container">
<div class="row">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body" role="main">
<section id="roadmap">
<span id="id1"></span><h1>Roadmap<a class="headerlink" href="#roadmap" title="Permalink to this headline">¶</a></h1>
<p>The following list of milestones is to guide the core developers on the future
direction of the package development. The list is by no means exhaustive and
will be updated over time as the development progresses and new algorithms
are proposed by the research community.</p>
<p>The list is algorithm- and feature-oriented as the goal of the package is to
give the community access to a tool that has all the necessary functionality
for FAT research and deployment.</p>
<section id="milestone-1">
<h2>Milestone 1 ✔<a class="headerlink" href="#milestone-1" title="Permalink to this headline">¶</a></h2>
<p>The first milestone is our first public release of the package – version
<em>0.0.1</em>. The following functionality should be available.</p>
<table class="docutils align-center">
<colgroup>
<col style="width: 14%" />
<col style="width: 29%" />
<col style="width: 28%" />
<col style="width: 28%" />
</colgroup>
<tbody>
<tr class="row-odd"><td></td>
<td><p>Fairness</p></td>
<td><p>Accountability</p></td>
<td><p>Transparency</p></td>
</tr>
<tr class="row-even"><td><p>Data/
Features</p></td>
<td><ul class="simple">
<li><p>Systemic Bias
(disparate treatment
labelling)</p></li>
<li><p>Sample size disparity
(e.g., class imbalance)</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>Sampling bias</p></li>
<li><p>Data Density Checker</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>Data description</p></li>
</ul>
</td>
</tr>
<tr class="row-odd"><td><p>Models</p></td>
<td><ul class="simple">
<li><p>Group-based fairness
(disparate impact)</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>Systematic performance
bias</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>Partial dependence</p></li>
<li><p>Individual conditional
expectation</p></li>
</ul>
</td>
</tr>
<tr class="row-even"><td><p>Predictions</p></td>
<td><ul class="simple">
<li><p>Counterfactual fairness
(disparate treatment)</p></li>
</ul>
</td>
<td></td>
<td><ul class="simple">
<li><p>Counterfactuals</p></li>
<li><p>Tabular LIME (wrapper)</p></li>
</ul>
</td>
</tr>
</tbody>
</table>
</section>
<section id="milestone-2">
<h2>Milestone 2<a class="headerlink" href="#milestone-2" title="Permalink to this headline">¶</a></h2>
<p>This will be the first major update of the package. The focus will be placed on
the transparency module. Nevertheless, some additional fairness and
accountability functionality will be implemented as well.</p>
<table class="docutils align-center">
<colgroup>
<col style="width: 14%" />
<col style="width: 29%" />
<col style="width: 28%" />
<col style="width: 28%" />
</colgroup>
<tbody>
<tr class="row-odd"><td></td>
<td><p>Fairness</p></td>
<td><p>Accountability</p></td>
<td><p>Transparency</p></td>
</tr>
<tr class="row-even"><td><p>Data/
Features</p></td>
<td></td>
<td><ul class="simple">
<li><p>k-anonymity</p></li>
<li><p>l-diversity</p></li>
<li><p>t-closeness</p></li>
</ul>
</td>
<td></td>
</tr>
<tr class="row-odd"><td><p>Models</p></td>
<td><ul class="simple">
<li><p>Additional fairness
metrics (to be
decided)</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>Background check</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>PD/ICE enhancements</p></li>
<li><p>Scikit-learn model
explainers</p></li>
<li><p>ANCHOR</p></li>
<li><p>Forestspy</p></li>
<li><p>Tree interpreter</p></li>
<li><p>Feature importance</p></li>
<li><p>Model reliance</p></li>
<li><p>TREPAN</p></li>
<li><p>Logical models
counterfactual
explainer
for and their
ensembles</p></li>
</ul>
</td>
</tr>
<tr class="row-even"><td><p>Predictions</p></td>
<td></td>
<td></td>
<td><ul class="simple">
<li><p>Scikit-learn
prediction
explainers</p></li>
<li><p>Generalised local
surrogates (bLIMEy)</p></li>
<li><p>bLIMEy LIME
implementation for
tabular, text and
image data</p></li>
</ul>
</td>
</tr>
</tbody>
</table>
<ul class="simple">
<li><p>Extra fairness metrics.</p>
<ul>
<li><p>Implement additional group-based fairness metrics.</p></li>
<li><p>Implement threshold computation based on the selected group metric
equality.</p></li>
<li><p>Implement Jupyter Notebook interactive plugins (widgets) to allow the
community to play with the fairness concepts. (E.g., widgets similar to
interactive figures in this <a class="reference external" href="https://research.google.com/bigpicture/attacking-discrimination-in-ml/">Google blog post</a>.</p></li>
</ul>
</li>
<li><p>Merge the pull request with k-anonimity, l-diversity and t-closeness.</p></li>
<li><p>Implement <a class="reference external" href="https://github.com/perellonieto/background_check/blob/master/jupyter/background_check.ipynb">Background Check</a>.</p></li>
<li><p>PD and ICE enhancements (pull request).</p>
<ul>
<li><p>2-D implementation.</p></li>
<li><p>Implementation for classification and regression.</p></li>
<li><p>Improved visualisations.</p></li>
</ul>
</li>
<li><p>Scikit-learn model explainers (cf. the reference implementation in the
<a class="reference external" href="https://eli5.readthedocs.io/en/latest/libraries/sklearn.html">eli5 package</a>).</p>
<ul>
<li><p>Decision trees.</p>
<ul>
<li><p>Feature importance.</p></li>
<li><p>Decision tree structure (tree plot).</p></li>
</ul>
</li>
<li><p>Rule lists and sets (these can share a common representation with the
trees).</p>
<ul>
<li><p>Rule list structure (rule list in a text form).</p></li>
</ul>
</li>
<li><p>Linear models.</p>
<ul>
<li><p>Feature importance (coefficients).</p></li>
</ul>
</li>
<li><p>K-means.</p>
<ul>
<li><p>Prototypes.</p>
<ul>
<li><p>Similarities between examples in a cluster that are correctly assigned
to this clusetr.</p></li>
</ul>
</li>
<li><p>Criticisms.</p>
<ul>
<li><p>Similarities between examples in a cluster that are incorrectly
assigned to this clusetr.</p></li>
</ul>
</li>
</ul>
</li>
</ul>
</li>
<li><p>Implement <a class="reference external" href="https://github.com/marcotcr/anchor">ANCHOR</a>.</p></li>
<li><p>Implement <a class="reference external" href="https://github.com/jvns/forestspy">forestspy</a>.</p></li>
<li><p>Implement <em>Tree Interpreter</em>.</p>
<ul>
<li><p>“The global feature importance of random forests can be quantified by the
total decrease in node impurity averaged over all trees of the ensemble
(‘mean decrease impurity’).”</p></li>
<li><p>“We can use the difference between the mean value of data points in a
parent node between that of a child node to approximate the contribution of
this split…”</p></li>
<li><p><a class="reference external" href="https://blog.datadive.net/interpreting-random-forests/">Interpreting random forests</a> and
<a class="reference external" href="https://blog.datadive.net/random-forest-interpretation-with-scikit-learn/">Random forest interpretation with scikit-learn</a> blog posts hold some
useful information extracted from the “Interpreting random forests” paper
by Ando Saabas.</p></li>
</ul>
</li>
<li><p>Implement a variety of feature importance metrics.</p>
<ul>
<li><p>Random forest feature (variable) importance (“Random Forests”, Leo Breiman,
2001). (Similar to <em>permutation importance</em>.)</p></li>
<li><p>XGboost feature importance.</p>
<ul>
<li><p>Feature weight – the number of times a feature appears in a tree
(ensemble).</p></li>
<li><p>Gain – the average gain of splits that use the feature.</p></li>
<li><p>Coverage – the average coverage (number of samples affected) of splits
that use the feature.</p></li>
</ul>
</li>
<li><p><a class="reference external" href="https://oracle.github.io/Skater/reference/interpretation.html#feature-importance">Skater feature importance</a>.</p></li>
<li><p>Prediction variance – mean absolute value of changes in predictions given
perturbations in the data.</p></li>
<li><p>“Variable Importance Analysis: A Comprehensive Review”. Reliability
Engineering & System Safety 142 (2015): 399-432; Wei, Pengfei, Zhenzhou Lu,
and Jingwen Song.</p></li>
<li><p>Scikit-learn and eli5 <strong>permutation importance</strong> (a.k.a.
<em>Mean Decrease Accuracy (MDA)</em>).</p>
<ul>
<li><p><a class="reference external" href="https://eli5.readthedocs.io/en/latest/autodocs/sklearn.html#module-eli5.sklearn.permutation_importance">eli5 implementation</a>.</p></li>
<li><p>(These may be sensitive to features being correlated – a user guide note
should suffice.)</p></li>
</ul>
</li>
</ul>
</li>
<li><p>Implement <em>model reliance</em> (Fisher, 2018). (“All Models are Wrong but many
are Useful: Variable Importance for Black-Box, Proprietary, or Misspecified
Prediction Models, using Model Class Reliance”, Aaron Fisher, Cynthia Rudin,
Francesca Dominici.)</p></li>
<li><p>Implement TREPAN (tree surrogate).</p>
<ul>
<li><p>“Extracting Comprehensible Models From Trained Neural Networks”, Mark W.
Craven(1996). (<a class="reference external" href="https://ftp.cs.wisc.edu/machine-learning/shavlik-group/craven.thesis.pdf">PhD thesis</a>)</p></li>
<li><p>“Extracting Thee-Structured Representations of Trained Networks”, Mark W.
Craven and Jude W. Shavlik (NIPS, 96). (<a class="reference external" href="https://papers.nips.cc/paper/1152-extracting-tree-structured-representations-of-trained-networks.pdf">NIPS paper</a>)</p></li>
<li><p>“Study of Various Decision Tree Pruning Methods with their Empirical
Comparison in WEKA”, Nikita Patel and Saurabh Upadhyay (2012). (<a class="reference external" href="https://pdfs.semanticscholar.org/025b/8c109c38dc115024e97eb0ede5ea873fffdb.pdf">report</a>)</p></li>
<li><p><a class="reference external" href="https://oracle.github.io/Skater/reference/interpretation.html#tree-surrogates-using-decision-trees">TREPAN implementation</a> in Skater.</p></li>
</ul>
</li>
<li><p>Implement a counterfactual explainer for logical models and their ensembles.</p></li>
<li><p>Scikit-learn prediction explainers.</p>
<ul>
<li><p>Decision trees.</p>
<ul>
<li><p>Root-to-leaf path (logical conditions).</p></li>
<li><p>Counterfactuals.</p></li>
</ul>
</li>
<li><p>Rule lists and sets.</p>
<ul>
<li><p>Logical conditions list (as text).</p></li>
</ul>
</li>
<li><p>Neighbours.</p>
<ul>
<li><p>Similarities and differences (on the feature vector) among the neighbours
of the same and the opposite class.</p></li>
</ul>
</li>
<li><p>K-means.</p>
<ul>
<li><p>Prototypes.</p>
<ul>
<li><p>Nearest centroid of the same class.</p></li>
</ul>
</li>
<li><p>Criticisms.</p>
<ul>
<li><p>Nearest centroid of the opposite class.</p></li>
</ul>
</li>
</ul>
</li>
</ul>
</li>
<li><p>bLIMEy implementation.</p></li>
<li><p>Fresh LIME implementation.</p>
<ul>
<li><p>Write tutorials similar to <a class="reference external" href="https://github.com/marcotcr/lime/tree/gh-pages/tutorials">LIME tutorials</a>, in particular
<a class="reference external" href="https://marcotcr.github.io/lime/tutorials/Lime%20-%20multiclass.html">this tutorial</a>.</p></li>
<li><p>Have a look at what eli5 does: “eli5.lime provides dataset generation
utilities for text data (remove random words) and for arbitrary data
(sampling using Kernel Density Estimation) … for explaining predictions
of probabilistic classifiers eli5 uses another classifier by default,
trained using cross-entropy loss, while canonical library fits regression
model on probability output.”</p></li>
</ul>
</li>
</ul>
</section>
<section id="milestone-3">
<h2>Milestone 3<a class="headerlink" href="#milestone-3" title="Permalink to this headline">¶</a></h2>
<p>The third milestone will integrate the tool with important machine learning and
fairness packages.</p>
<table class="docutils align-center">
<colgroup>
<col style="width: 14%" />
<col style="width: 29%" />
<col style="width: 28%" />
<col style="width: 28%" />
</colgroup>
<tbody>
<tr class="row-odd"><td></td>
<td><p>Fairness</p></td>
<td><p>Accountability</p></td>
<td><p>Transparency</p></td>
</tr>
<tr class="row-even"><td><p>Data/
Features</p></td>
<td></td>
<td></td>
<td></td>
</tr>
<tr class="row-odd"><td><p>Models</p></td>
<td><ul class="simple">
<li><p>Fairness360 integration</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>Distribution shift
detection</p></li>
<li><p>Calibration</p></li>
</ul>
</td>
<td><ul class="simple">
<li><p>SHAP package
integration (Shapley
sampling values &
Shapley regression
values)</p></li>
<li><p>Xgboost package
interpreter</p></li>
<li><p>LightGBM package
interpreter</p></li>
<li><p>Lightning package
interpreter</p></li>
<li><p>Sklearn-crfsuite
package interpreter</p></li>
<li><p>eli5 package
integration</p></li>
<li><p>Bayesian Rule Lists
(BRL)</p></li>
<li><p>PD/ICE speed
improvements</p></li>
<li><p>Interactive (JS)
Jupyter Notebook plots</p></li>
</ul>
</td>
</tr>
<tr class="row-even"><td><p>Predictions</p></td>
<td></td>
<td></td>
<td><ul class="simple">
<li><p>SHAP package
integration</p></li>
<li><p>Xgboost package
interpreter</p></li>
</ul>
</td>
</tr>
</tbody>
</table>
<ul class="simple">
<li><p>Integration or reimplementation of <a class="reference external" href="https://github.com/IBM/AIF360">fairness360</a> package (depending on the</p></li>
<li><p>code quality).</p></li>
<li><p>Implement <em>distribution shift</em> metrics.</p></li>
<li><p>Implement <em>calibration</em> techniques.</p></li>
<li><p>Integration with the <a class="reference external" href="https://github.com/slundberg/shap">SHAP</a> package.</p></li>
<li><p>Explainers for models implemented in the <a class="reference external" href="https://github.com/dmlc/xgboost">Xgboost</a> package.</p></li>
<li><p>Explainers for models implemented in the <a class="reference external" href="https://github.com/microsoft/LightGBM">LightGBM</a> package.</p></li>
<li><p>Explainers for models implemented in the <a class="reference external" href="https://github.com/scikit-learn-contrib/lightning">Lightning</a> package.</p></li>
<li><p>Explainers for models implemented in the <a class="reference external" href="https://github.com/TeamHG-Memex/sklearn-crfsuite">sklearn-crfsuite</a> package.</p></li>
<li><p><a class="reference external" href="https://eli5.readthedocs.io/en/latest/libraries/index.html">eli5</a> integration. (“Text processing utilities from scikit-learn and can
highlight text data accordingly. Pipeline and FeatureUnion are supported. It
also allows to debug scikit-learn pipelines which contain HashingVectorizer,
by undoing hashing.”)</p></li>
<li><p>Implement Bayesian Rule Lists (BRL).</p>
<ul>
<li><p>Bayesian Rule Lists (BRL).</p>
<ul>
<li><p><a class="reference external" href="https://oracle.github.io/Skater/reference/interpretation.html#bayesian-rule-lists-brl">BRL reference implementation</a>.</p></li>
<li><p>Example <a class="reference external" href="https://arxiv.org/abs/1511.01644">BRL use case</a>: “Interpretable classifiers using rules and
Bayesian analysis: Building a better stroke prediction model”, Letham
et.al(2015).</p></li>
</ul>
</li>
<li><p>Scalable Bayesian Rule Lists (SBRL).</p>
<ul>
<li><p>“Scalable Bayesian Rule Lists”, Yang et.al (2016). (<a class="reference external" href="https://arxiv.org/abs/1602.08610">SBRL paper</a>)</p></li>
<li><p>Bayesian Rule List Classifier (<a class="reference external" href="https://github.com/Hongyuy/sbrl-python-wrapper/blob/master/sbrl/C_sbrl.py">BRLC</a>) is a Python wrapper for the SBRL.</p></li>
</ul>
</li>
<li><p>Big Data Bayesian Rule List Classifier (BigDataBRLC) is a BRLC to handle
large data-sets.</p>
<ul>
<li><p>Skater’s <a class="reference external" href="https://oracle.github.io/Skater/reference/interpretation.html#skater.core.global_interpretation.interpretable_models.bigdatabrlc.BigDataBRLC">BigDataBRLC implementation</a>.</p></li>
<li><p><a class="reference external" href="https://github.com/tmadl/sklearn-expertsys/blob/master/BigDataRuleListClassifier.py">Dr. Tamas Madl’s implementation</a>.</p></li>
</ul>
</li>
</ul>
</li>
<li><p>PD/ICE speed improvements – parallelisation and a progress bar.</p></li>
<li><p>iPython/Jupyter Notebook interactive (JS) plots to improve research
applicability aspect of the package.</p></li>
</ul>
</section>
<section id="milestone-4">
<h2>Milestone 4<a class="headerlink" href="#milestone-4" title="Permalink to this headline">¶</a></h2>
<p>This milestone is focused on implementing in the package a collection of tools
that will enable researchers and practitioners to use it with (deep) neural
networks (Deep Learning, autograd, optimisation).</p>
<table class="docutils align-center">
<colgroup>
<col style="width: 14%" />
<col style="width: 29%" />
<col style="width: 28%" />
<col style="width: 28%" />
</colgroup>
<tbody>
<tr class="row-odd"><td></td>
<td><p>Fairness</p></td>
<td><p>Accountability</p></td>
<td><p>Transparency</p></td>
</tr>
<tr class="row-even"><td><p>Data/
Features</p></td>
<td></td>
<td></td>
<td></td>
</tr>
<tr class="row-odd"><td><p>Models</p></td>
<td><ul class="simple">
<li><p>what-if tool
integration</p></li>
</ul>
</td>
<td></td>
<td><ul class="simple">
<li><p>Quantitative Input
influence (QII)</p></li>
<li><p>Layer-wise Relevance
Propagation (e-LRP)</p></li>
<li><p>Occlusion</p></li>
<li><p>integrated gradient</p></li>
<li><p>what-if tool
integration</p></li>
</ul>
</td>
</tr>
<tr class="row-even"><td><p>Predictions</p></td>
<td></td>
<td></td>
<td><ul class="simple">
<li><p>DeepLIFT (example
explanation)</p></li>
<li><p>DeepExplain</p></li>
</ul>
</td>
</tr>
</tbody>
</table>
<ul class="simple">
<li><p>Integration with the <a class="reference external" href="https://pair-code.github.io/what-if-tool/">what-if tool</a>.</p></li>
<li><p>Implement Quantitative Input influence (QII).</p></li>
<li><p>Implement epsilon-Layer-wise Relevance Propagation (e-LRP).</p>
<ul>
<li><p>“On Pixel-Wise Explanations for Non-Linear Classifier Decisions by
Layer-Wise Relevance Propagation”, Bach S, Binder A, Montavon G, Klauschen
F, Muller K-R, Samek W (2015).</p></li>
<li><p>“Towards better understanding of gradient-based attribution methods for
Deep Neural Networks”, Ancona M, Ceolini E, Oztireli C, Gross M (ICLR,
2018).</p></li>
</ul>
</li>
<li><p>Implement <em>occlusion</em>.</p>
<ul>
<li><p>“Visualizing and understanding convolutional networks”, Zeiler, M and
Fergus, R (Springer, 2014).</p></li>
<li><p><a class="reference external" href="https://github.com/marcoancona/DeepExplain/blob/master/deepexplain/tensorflow/methods.py">Occlusion implementation</a>.</p></li>
</ul>
</li>
<li><p>Implement Integrated Gradient method.</p>
<ul>
<li><p>“Axiomatic Attribution for Deep Networks”, Sundararajan, M, Taly, A, Yan, Q
(ICML, 2017).</p></li>
<li><p><a class="reference external" href="https://theory.stanford.edu/~ataly/Talks/sri_attribution_talk_jun_2017.pdf">Integrated Gradient slides</a>.</p></li>
</ul>
</li>
<li><p>Implement the DeepLIFT algorithm.</p></li>
<li><p>Implement the DeepExplain algorithm.</p>
<ul>
<li><p>“Towards better understanding of gradient-based attribution methods for
Deep Neural Networks”, Ancona M, Ceolini E, Oztireli C, Gross M
(ICLR, 2018).</p></li>
<li><p><a class="reference external" href="https://github.com/marcoancona/DeepExplain/blob/master/deepexplain/tensorflow/methods.py">DeepExplain implementation</a>.</p></li>
</ul>
</li>
</ul>
<hr class="docutils" />
<ul class="simple">
<li><p>Finalise full integration of Skater and SHAP (deep neural netowrks).</p></li>
</ul>
</section>
</section>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
<div class="container cusotm-footer">
<div class="row">
<div class="col-3">
© 2018--2023, Kacper Sokol et al.
</div>
<div class="col-6">
<a class="footer-img-link" href="contributors.html#funding">
<img src="_static/img/bristol.svg" title="Univeristy of Bristol" style="max-height: 30px">
<img src="_static/img/thales.svg" title="Thales" style="max-height: 20px">
</a>
<br>
<a href="contributors.html#funding" style="top: 4px; position: relative;">
More information on our contributors
</a>
</div>
<div class="col-3">
<a href="_sources/roadmap.rst.txt" rel="nofollow">
Show this page source
</a>
</div>
</div>
</div>
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-FHEM8Y8CHX"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-FHEM8Y8CHX');
</script>
<script>
(function() {
var cx = '014039732624725851926:mpducezglrq';
var gcse = document.createElement('script');
gcse.type = 'text/javascript';
gcse.async = true;
gcse.src = 'https://cse.google.com/cse.js?cx=' + cx;
var s = document.getElementsByTagName('script')[0];
s.parentNode.insertBefore(gcse, s);
})();
</script>
</body>
</html>