merge conflict

README.md

@@ -1,6 +1,6 @@
-# Georgia Tech Borglab Website
+# OPTML@Michigan State University Website
 
-This is the website of Georgia Tech's Borglab, led by [Frank Dellaert](https://dellaert.hithub.io)
+This is the website of Michigan State University's OPTML Lab, led by [Sijia Liu]([https://dellaert.hithub.io](https://lsjxjtu.github.io/))
 
 It was forked from the Leiden University's Allen Lab repo, powered by Jekyll and some Bootstrap, Bootwatch. Go to *aboutwebsite.md*  to learn how to copy and modidy this page for your purpose. 
 

_data/alumni_visitors.yml

@@ -1,25 +1,25 @@
-- name: Chongyu Fan (remote)
+- name: "<a href='https://a-f1.github.io/'>Chongyu Fan</a>"
   info: Undergraduate, Huazhong University of Science and Technology, China
   visiting: May. 2023 - Present
-  research: "<a href='https://arxiv.org/abs/2310.12508'>SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation</a>"
+  research: "<a href='https://arxiv.org/abs/2310.12508'>SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation</a> [<b>Spotlight</b> paper at ICLR’24]"
 
 - name: Zhuoshi Pan (remote)
   info: Graduate student, Tsinghua University, China
   visiting: May. 2023 - Present
-  research: "<a href='https://arxiv.org/pdf/2311.02373v1.pdf'>From Trojan Horses to Castle Walls: Unveiling Bilateral Backdoor Effects in Diffusion Models</a>"
+  research: "<a href='https://arxiv.org/abs/2311.02373'>From Trojan Horses to Castle Walls: Unveiling Bilateral Backdoor Effects in Diffusion Models</a> [Paper at <a href='https://neurips2023-bugs.github.io/'>Backdoors in Deep Learning@NeurIPS’23</a>]"
 
 - name: Chen Fan (remote)
   info: Graduate student, University of British Columbia, Canada
   visiting: May 2021 - Sept. 2021
-  research: "<a href='https://arxiv.org/abs/2109.07497'>Sign-MAML: Efficient Model-Agnostic Meta-Learning by SignSGD</a> [Paper at <a href='https://meta-learn.github.io/2021/'>MetaLearn@NeurIPS'21</a>]"
+  research: "<a href='https://arxiv.org/abs/2109.07497'>Sign-MAML: Efficient Model-Agnostic Meta-Learning by SignSGD</a> [Paper at <a href='https://meta-learn.github.io/2021/'>MetaLearn@NeurIPS’21</a>]"
 - name: Vardaan Taneja (remote)
   info: Undergraduate, Indian Institute of Technology, Delhi, India
   visiting: May 2021 - Sept. 2021
   research: "<a href='https://ieeexplore.ieee.org/document/9746433'>When Does Backdoor Attack Succeed in Image Reconstruction? A Study of Heuristics vs. Bi-Level Solution</a> [Paper at ICASSP’22]"
 - name: "<a href='https://scholar.google.ca/citations?user=c2VU-_4AAAAJ&hl=en'>Soumyadeep Pal</a>"
   info: Graduate student, University of Alberta, Canada
   visiting: May 2022 - Sept. 2022
-  research: "<a href='https://arxiv.org/abs/2301.08751'>Towards Understanding How Self Training Tolerates Data Backdoor Poisoning</a> [Best Paper Award Finalist, <a href='https://safeai.webs.upv.es/index.php/best-paper-award/?'>SafeAI Workshop@AAAI'23</a>]"
+  research: "<a href='https://arxiv.org/abs/2301.08751'>Towards Understanding How Self Training Tolerates Data Backdoor Poisoning</a> [Best Paper Award Finalist, <a href='https://safeai.webs.upv.es/index.php/best-paper-award/?'>SafeAI Workshop@AAAI’23</a>]"
 - name: "<a href='https://scholar.google.at/citations?user=sb4hPQMAAAAJ&hl=en'>Peter Lorenz</a> (remote)"
   info: Graduate student, Fraunhofer Center of Machine Learning, Germany
   visiting: May 2022 - Sept. 2022
@@ -31,4 +31,4 @@
 - name: Haomin Zhuang (remote)
   info: Undergraduate, South China University of Technology, China
   visiting: Oct. 2022 - Mar. 2023
-  research: "<a href='https://arxiv.org/abs/2303.16378'>A Pilot Study of Query-Free Adversarial Attack against Stable Diffusion</a> [Paper at <a href='https://robustart.github.io/'>The 3rd Workshop of Adversarial Machine Learning on Computer Vision@CVPR23</a>]"
+  research: "<a href='https://arxiv.org/abs/2303.16378'>A Pilot Study of Query-Free Adversarial Attack against Stable Diffusion</a> [Paper at <a href='https://robustart.github.io/'>The 3rd Workshop of Adversarial Machine Learning on Computer Vision@CVPR’23</a>]"

_data/preprints.yml

@@ -11,13 +11,20 @@
   post:
 
 - title: "To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Now"
-  image: dummy.png
-  description:
+  image: mu_attack.png
+  description: "The recent advances in diffusion models (DMs) have revolutionized the generation of complex and diverse images. However, these models also introduce potential safety hazards, such as the production of harmful content and infringement of data copyrights. Although there have been efforts to create safety-driven unlearning methods to counteract these challenges, doubts remain about their capabilities. To bridge this uncertainty, we propose an evaluation framework built upon adversarial attacks (also referred to as adversarial prompts), in order to discern the trustworthiness of these safety-driven unlearned DMs. Specifically, our research explores the (worst-case) robustness of unlearned DMs in eradicating unwanted concepts, styles, and objects, assessed by the generation of adversarial prompts. We develop a novel adversarial learning approach called UnlearnDiff that leverages the inherent classification capabilities of DMs to streamline the generation of adversarial prompts, making it as simple for DMs as it is for image classification attacks. This technique streamlines the creation of adversarial prompts, making the process as intuitive for generative modeling as it is for image classification assaults. Through comprehensive benchmarking, we assess the unlearning robustness of five prevalent unlearned DMs across multiple tasks. Our results underscore the effectiveness and efficiency of UnlearnDiff when compared to state-of-the-art adversarial prompting methods."
   authors:
     "<span style='color:blue'>Y. Zhang*</span>, <span style='color:blue'>J. Jia*</span>, X. Chen, <span style='color:blue'>A. Chen</span>, <span style='color:blue'>Y. Zhang</span>, <span style='color:blue'>J. Liu</span>, K. Ding, <span style='color:blue'>S. Liu</span>"
   link:
     url: https://arxiv.org/abs/2310.11868
+<<<<<<< HEAD
     display: 
   highlight: 0
   news2: 
   post:
+=======
+    display: Under Review
+  highlight: 1
+  news2: 
+  post: mu_attack
+>>>>>>> 335352f3ef3081bcc89f132f46da88c8d95425a5

_data/publist.yml

@@ -5,23 +5,22 @@
     "<span style='color:blue'>C. Fan*</span>, <span style='color:blue'>J. Liu*</span>, <span style='color:blue'>Y. Zhang</span>, D. Wei, E. Wong, <span style='color:blue'>S. Liu</span>"
   link:
     url: https://arxiv.org/abs/2310.12508
-    display: ICLR'24 Spotlight
+    display: ICLR'24 (<span style='color:red'><b>Spotlight</b>, acceptance rate 5%</span>)
   highlight: 0
   news2: 
   post:
 
 - title: "DeepZero: Scaling up Zeroth-Order Optimization for Deep Model Training"
-  image: dummy.png
-  description:
+  image: deepzero_iclr24.png
+  description: "Zeroth-order (ZO) optimization has become a popular technique for solving machine learning (ML) problems when first-order (FO) information is difficult or impossible to obtain. However, the scala- bility of ZO optimization remains an open problem: Its use has primarily been limited to relatively small-scale ML problems, such as sample-wise adversarial attack generation. To our best knowl- edge, no prior work has demonstrated the effectiveness of ZO optimization in training deep neural networks (DNNs) without a significant decrease in performance. To overcome this roadblock, we develop DeepZero, a principled ZO deep learning (DL) framework that can scale ZO optimization to DNN training from scratch through three primary innovations. First, we demonstrate the advantages of coordinate-wise gradient estimation (CGE) over randomized vector-wise gradient estimation in training accuracy and computational efficiency. Second, we propose a sparsity-induced ZO training protocol that extends the model pruning methodology using only finite differences to explore and exploit the sparse DL prior in CGE. Third, we develop the methods of feature reuse and forward parallelization to advance the practical implementations of ZO training. Our extensive experiments show that DeepZero achieves state-of-the-art (SOTA) accuracy on ResNet-20 trained on CIFAR-10, approaching FO training performance for the first time. Furthermore, we show the practical utility of DeepZero in applications of certified adversarial defense and DL-based partial differential equa- tion error correction, achieving 10-20% improvement over SOTA. We believe our results will inspire future research on scalable ZO optimization and contribute to advancing DL with black box."
   authors:
     "<span style='color:blue'>A. Chen*</span>, <span style='color:blue'>Y. Zhang*</span>, <span style='color:blue'>J. Jia</span>, J. Diffenderfer, <span style='color:blue'>J. Liu</span>, K. Parasyris, <span style='color:blue'>Y. Zhang</span>, Z. Zhang, B. Kailkhura, <span style='color:blue'>S. Liu</span>"
   link:
     url: https://arxiv.org/abs/2310.02025
     display: ICLR'24
-  highlight: 0
+  highlight: 1
   news2: 
-  post:
-
+  post: deepzero_iclr24
 
 - title: "AutoVP: An Automated Visual Prompting Frameowrk and Benchmark"
   image: dummy.png
@@ -49,16 +48,16 @@
   post:
 
 - title: "Model Sparsity Can Simplify Machine Unlearning"
-  image: dummy.png
-  description:
+  image: Sparse_unlearn_neurips23.png
+  description: "Machine unlearning (MU) has become essential for complying with data regulations by eliminating the influence of specific data from models. Traditional exact unlearning methods, which involve retraining from scratch, are computationally expensive, prompting the exploration of efficient, approximate alternatives. Our research introduces a model-based approach: sparsity through weight pruning,that narrows the gap between exact and approximate unlearning. We present a new paradigm, “prune first, then unlearn,” which integrates a sparsity model into unlearning, and a sparsity-aware technique that further refines approximate unlearning training. Our extensive experiments confirm the effectiveness of our methods, particularly a 77% increase in unlearning efficacy with fine-tuning, and their applicability in mitigating backdoor attacks and improving transfer learning."
   authors:
     "<span style='color:blue'>J. Jia*</span>, <span style='color:blue'>J. Liu*</span>, P. Ram, <span style='color:blue'>Y. Yao</span>, G. Liu, Y. Liu, P. Sharma, <span style='color:blue'>S. Liu</span>"
   link:
     url: https://arxiv.org/abs/2304.04934
-    display: NeurIPS'23 (<span style='color:red'><b>Spotlight</b></span>)
-  highlight: 0
+    display: NeurIPS'23 (<span style='color:red'><b>Spotlight</b>, acceptance rate 3%</span>)
+  highlight: 1
   news2: 
-  post:
+  post: Sparse_unlearn_neurips23
 
 - title: "On the Convergence and Sample Complexity Analysis of Deep Q-Networks with Greedy Exploration"
   image: dummy.png
@@ -79,7 +78,7 @@
     "<span style='color:blue'>Y. Zhang</span>, R. Cai, T. Chen, G. Zhang, H. Zhang. P. Chen, S. Chang, Z. Wang, <span style='color:blue'>S. Liu</span>"
   link:
     url: https://arxiv.org/abs/2308.10110v1
-    display: ICCV'23 (<span style='color:red'><b>Oral</b></span>)
+    display: ICCV'23 (<span style='color:red'><b>Oral</b>, acceptance rate 2%</span>)
   highlight: 0
   news2: 
   post:

_pages/404.md

@@ -1,7 +1,7 @@
 ---
 title: "Page Not Found"
 layout: default
-excerpt: "Page not found. Your pixels are in another canvas."
+# excerpt: "Page not found. Your pixels are in another canvas."
 sitemap: false
 permalink: /404.html
 ---

_pages/aboutwebsite.md

@@ -1,7 +1,7 @@
 ---
 title: "About the website"
 layout: textlay
-excerpt: "About the website."
+# excerpt: "About the website."
 sitemap: false
 permalink: /aboutwebsite.html
 ---

_pages/allnews.md

@@ -1,7 +1,7 @@
 ---
 title: "News"
 layout: textlay
-excerpt: "OPTML at Michigan State University"
+# excerpt: "OPTML at Michigan State University"
 sitemap: false
 permalink: /allnews.html
 ---

_pages/home.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML - Home"
 layout: homelay
-excerpt: "OPTML Group at Michigan State University."
+# excerpt: "OPTML Group at Michigan State University."
 sitemap: false
 permalink: /
 ---

_pages/openings.md

@@ -1,7 +1,7 @@
 ---
 title: "Joining the OPTML Group"
 layout: textlay
-excerpt: "Openings"
+# excerpt: "Openings"
 sitemap: false
 permalink: /vacancies
 ---

_pages/pictures.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML - Photos"
 layout: piclay
-excerpt: "OPTML -- Photos"
+# excerpt: "OPTML -- Photos"
 permalink: /pictures/
 ---
 

_pages/pictures_old.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML - Photos"
 layout: piclay
-excerpt: "OPTML -- Photos"
+# excerpt: "OPTML -- Photos"
 permalink: /pictures_old/
 ---
 

_pages/posts.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML Group - Projects"
 layout: gridlay
-excerpt: "OPTML Group -- Projects"
+# excerpt: "OPTML Group -- Projects"
 sitemap: false
 permalink: /posts
 ---
@@ -11,6 +11,37 @@ permalink: /posts
 Authors marked in <span style="color:blue">blue</span> indicate our group members, and "\*" indicates equal contribution.<br />
 
 {% assign number_printed = 0 %}
+{% for publi in site.data.preprints %}
+
+{% assign even_odd = number_printed | modulo: 2 %}
+{% if publi.highlight == 1 %}
+
+{% if even_odd == 0 %}
+<div class="row">
+{% endif %}
+
+<div class="col-sm-6 clearfix">
+ <div class="well">
+  <pubtit>{{ publi.title }}</pubtit>
+  <img src="{{ site.url }}{{ site.baseurl }}/images/pubpic/{{ publi.image }}" class="img-responsive" width="100%" style="float: left" />
+  <p>{{ publi.description }}</p>
+  <p><em>{{ publi.authors }}</em></p>
+  <p><strong><a href="{{ publi.link.url }}">{{ publi.link.display }}</a></strong></p>
+  <p class="text-danger"><strong> {{ publi.news1 }}</strong></p>
+  <p> {{ publi.news2 }}</p>
+  <p><a href="{{ site.url }}{{ site.baseurl }}/posts/{{ publi.post }}">Click here to read the full article.</a></p>
+ </div>
+</div>
+
+{% assign number_printed = number_printed | plus: 1 %}
+
+{% if even_odd == 1 %}
+</div>
+{% endif %}
+
+{% endif %}
+{% endfor %}
+
 {% for publi in site.data.publist %}
 
 {% assign even_odd = number_printed | modulo: 2 %}

_pages/publications.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML Group - Publications"
 layout: gridlay
-excerpt: "OPTML Group -- Publications."
+# excerpt: "OPTML Group -- Publications."
 sitemap: false
 permalink: /publications
 ---

_pages/research.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML - Research"
 layout: textlay
-excerpt: "OPTML -- Research"
+# excerpt: "OPTML -- Research"
 sitemap: false
 permalink: /research
 ---

_pages/team.md

@@ -1,7 +1,7 @@
 ---
 title: "OPTML - Team"
 layout: gridlay
-excerpt: "OPTML: Team members"
+# excerpt: "OPTML: Team members"
 sitemap: false
 permalink: /team/
 ---

_pages/twitterfeed.md

@@ -1,7 +1,7 @@
 ---
 title: "Timeline"
 layout: textlay
-excerpt: "Borglab at Georgia Tech."
+# excerpt: "Borglab at Georgia Tech."
 sitemap: false
 permalink: 
 ---