<div dir="ltr">Team,<div><br></div><div>Please consider attending a thesis proposal presentation by Nick (this Tuesday, ie. tomorrow, 10am, see zoom link below).</div><div>This talk looks to be really fun to hear.</div><div><br></div><div>Cheers</div><div>Artur<br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">---------- Forwarded message ---------<br>From: <strong class="gmail_sendername" dir="auto">Suzanne Lyons Muth</strong> <span dir="auto"><<a href="mailto:lyonsmuth@cmu.edu">lyonsmuth@cmu.edu</a>></span><br>Date: Mon, Jan 11, 2021 at 9:40 AM<br>Subject: RI Ph.D. Thesis Proposal: Nicholas Gisolfi<br>To: <a href="mailto:ri-people@lists.andrew.cmu.edu">ri-people@lists.andrew.cmu.edu</a> <<a href="mailto:ri-people@lists.andrew.cmu.edu">ri-people@lists.andrew.cmu.edu</a>><br></div><br><br>
<div dir="ltr">
<div id="m_5272019604846404083divtagdefaultwrapper" style="font-size:11pt;color:#000000;font-family:'Calibri Light','Helvetica Light',sans-serif" dir="ltr">
<p style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont">
<font style="color:rgb(33,33,33);font-size:15px;font-family:arial,sans-serif,serif,EmojiFont"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont;font-size:11pt"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Date:
19 </span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">January 2021</span></span><br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont;font-size:11pt"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Time: </span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">10:00
a</span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">.m. (ET)</span></span><br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Place: *Virtual Presentation* </span></font><span style="color:rgb(33,33,33);font-family:wf_segoe-ui_normal,"Segoe UI","Segoe WP",Tahoma,Arial,sans-serif,serif,EmojiFont;font-size:15px"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt"> </span></span><span style="color:rgb(33,33,33);font-family:wf_segoe-ui_normal,"Segoe UI","Segoe WP",Tahoma,Arial,sans-serif,serif,EmojiFont;font-size:15px"></span><a href="https://cmu.zoom.us/j/96434017510?pwd=amJXYlFTUDAvMmp6NkprUUtxR0xoUT09" rel="noopener noreferrer" style="font-family:wf_segoe-ui_normal,"Segoe UI","Segoe WP",Tahoma,Arial,sans-serif,serif,EmojiFont;font-size:15px" id="m_5272019604846404083LPlnk61192" target="_blank"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">https://cmu.zoom.us/j/96434017510?pwd=amJXYlFTUDAvMmp6NkprUUtxR0xoUT09</span></a></p>
<div style="color:rgb(33,33,33)"><font style="font-family:arial,sans-serif,serif,EmojiFont;font-size:15px"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Type: Ph.D. Thesis Proposal</span><br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Who: Nicholas Gisolfi</span><br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Title: </span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Verification and Accreditation of Artificial Intelligence</span><br>
<br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Abstract</span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">:</span><br>
</font>
<div style="font-family:wf_segoe-ui_normal,"Segoe UI","Segoe WP",Tahoma,Arial,sans-serif,serif,EmojiFont;font-size:15px">
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont">
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">This work involves formally verifying a trained model's adherence to important design specifications for the purpose of model accreditation. Accreditation of a trained
model requires enumeration of the explicit operational conditions under which the model is certified to meet all necessary specifications. By verifying model adherence to specifications set by developers, we increase the trustworthiness of the model along
the dimensions that are most relevant to earning a developer's trust. We argue that this gives developers a tool to quantitatively define and formally answer the fundamental question of ‘should we trust this model’.</span></div>
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont;min-height:15px">
<br>
</div>
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont">
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">We intend to demonstrate utility of the framework with the verification of a repertoire of design specifications on voting tree ensemble models for classification.
Verification is posed as an instance of the Boolean Satisfiability (SAT) problem. Our novel SAT encoding yields disruptive speed gains over related tree ensemble verification techniques, which facilitates the verification of harder specifications on models
of larger scale than reported in literature. It is possible to extend our framework to other model classes and design specifications that are beyond our current scope.</span></div>
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont;min-height:15px">
<br>
</div>
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont">
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">We show it is possible to use our framework to provide explanations for model behavior. We introduce a quantitative definition for interpretability; a provably correct
interpretation of a model is a causal specification relating inputs to outputs that the model never violates. We demonstrate how we can mine data for a specific type of causal structure; 2D, axis-aligned range rules that offer candidate specifications for
verification. These bounding box rules represent intelligible, intermediate-level concepts that, pending verification, offer provably correct summaries of model behavior. In addition to explanations for input-output behavior of a model, we demonstrate that
it is possible to provide explanations that diagnose discrepancies between what is observed and what was prescribed; if the model violates a required specification, we want to understand why. We provide explanations by summarizing proofs of unsatisfiability
that accompany the certificates. A minimized unsatisfiable set comprises the literals and clauses implicated in a logical contradiction. These provide the basis for provably correct explanations that are counterfactual and contrastive in nature.</span></div>
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont;min-height:15px">
<br>
</div>
<div style="color:rgb(0,0,0);margin:0px;font-size:13px;line-height:normal;font-family:"Helvetica Neue",serif,EmojiFont">
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">We propose a few directions to further demonstrate and expand the utility of our accreditation framework. For a supervised radiation risk assessment application, where
safety is paramount, we propose verifying a suite of new safety criteria that will increase domain-expert trust in the model's predictions. For selected clinical decision support applications, we propose to diagnose the sources of systematic disagreements
between opinions of experts who annotate training data, and to assess the model's adherence to common-sense-based expectations of its behavior. We expect these new capabilities to enhance adoption and effectiveness of artificial intelligence applications
to guide clinical decisions.</span></div>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont;font-size:11pt"></span></div>
<div><br>
<span style="font-family:arial,sans-serif,serif,EmojiFont;font-size:15px"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Thesis Committee Members</span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">:</span></span><br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont;font-size:11pt"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Artur Dubrawski</span><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">,
Chair</span></span><br>
<font face="Calibri Light, Helvetica Light, sans-serif, serif, EmojiFont"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Stephen Smith</span></font></div>
<div><font face="Calibri Light, Helvetica Light, sans-serif, serif, EmojiFont"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Reid Simmons</span></font></div>
<div><span style="color:rgb(0,0,0);font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">Madalina Fiterau, University of Massachusetts Amherst</span><font face="Calibri Light, Helvetica Light, sans-serif, serif, EmojiFont"><br>
</font><br>
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">A copy of the thesis proposal document is available at:</span></div>
<div><a href="https://cmu.box.com/s/gq9lumhvw7r0cnug3t9jnzrrb0jfcg5j" rel="noopener noreferrer" style="font-family:wf_segoe-ui_normal,"Segoe UI","Segoe WP",Tahoma,Arial,sans-serif,serif,EmojiFont;font-size:15px" title="https://cmu.box.com/s/gq9lumhvw7r0cnug3t9jnzrrb0jfcg5j
Cmd+Click or tap to follow the link" id="m_5272019604846404083LPlnk830820" target="_blank"><span style="font-family:"Calibri Light","Helvetica Light",sans-serif;font-size:11pt">https://cmu.box.com/s/gq9lumhvw7r0cnug3t9jnzrrb0jfcg5j</span></a><span style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont;font-size:14.666666984558105px"><br>
</span></div>
<div style="font-family:wf_segoe-ui_normal,"Segoe UI","Segoe WP",Tahoma,Arial,sans-serif,serif,EmojiFont;font-size:15px">
<span style="font-family:"Calibri Light","Helvetica Light",sans-serif,serif,EmojiFont;font-size:11pt"></span></div>
</div>
</div>
</div>
_______________________________________________<br>
ri-people mailing list<br>
<a href="mailto:ri-people@lists.andrew.cmu.edu" target="_blank">ri-people@lists.andrew.cmu.edu</a><br>
<a href="https://lists.andrew.cmu.edu/mailman/listinfo/ri-people" rel="noreferrer" target="_blank">https://lists.andrew.cmu.edu/mailman/listinfo/ri-people</a><br>
</div></div></div>