adding dtpinn paper and cv

Author: rsmath

.jekyll-cache/Jekyll/Cache/Jekyll--Converters--Markdown/d0/13dc87159a0c22d04a3455bb15f54ba17a6f9192baf26fba6e6820a0a8dba5

@@ -65,7 +65,8 @@ <h1>
                 <a href="https://scholar.google.com/citations?user=lUmqHckAAAAJ&hl=en">Google Scholar</a> &nbsp;/&nbsp;
                 <a href="https://github.com/ramanshsharma2806">GitHub</a> &nbsp;/&nbsp;
                 <a href="https://twitter.com/ramanshsharma1">Twitter</a> &nbsp;/&nbsp;
-                <a href="https://www.linkedin.com/in/ramanshsharma">LinkedIn</a>
+                <a href="https://www.linkedin.com/in/ramanshsharma">LinkedIn</a> &nbsp;/&nbsp;
+                <a href="/docs/Ramansh_Sharma_Resume.pdf">CV</a>
               </p>
             </td>
             <td style="padding:2.5%;width:40%;max-width:40%">

.jekyll-metadata

@@ -6,7 +6,7 @@ image: /images/kddpaper.png
 categories: research
 author: "Ramansh Sharma"
 authors: "Jordi Planas, Daniel Firebanks-Quevedo, Galina Naydenova, <strong>Ramansh Sharma</strong>, Cristina Taylor, Kathleen Buckingham, Rong Fang"
-venue: "Fragile Earth proceedings - KDD 2021"
+venue: "Fragile Earth proceedings - KDD"
 paper: /docs/kddpaper.pdf
 code: https://github.com/wri-dssg-omdena/policy-data-analyzer
 arxiv: https://arxiv.org/abs/2201.07105

_layouts/default.html

@@ -0,0 +1,15 @@
+---
+layout: post
+title:  "Accelerated Training of Physics Informed Neural Networks (PINNs) using Meshless Discretizations"
+date:   2022-09-15 01:59:59 +00:00
+image: /images/dtpinn.png
+categories: research
+author: "Ramansh Sharma"
+authors: "<strong>Ramansh Sharma</strong>, Varun Shankar"
+venue: "Neural Information Processing Systems (NeurIPS)"
+paper: /docs/dtpinn.pdf
+code: https://github.com/ramanshsharma2806/dt-pinn
+arxiv: https://arxiv.org/abs/2205.09332
+---
+
+We present a new technique for the accelerated training of physics-informed neural networks (PINNs): discretely-trained PINNs (DT-PINNs). DT-PINNs are trained by replacing exact spatial derivatives with high-order accurate numerical discretizations computed using meshless radial basis function-finite differences (RBF-FD) and applied via sparse-matrix vector multiplication. Additionally, though traditional PINNs (vanilla-PINNs) are typically stored and trained in 32-bit floating-point (fp32) on the GPU, we show that for DT-PINNs, using fp64 on the GPU leads to significantly faster training times than fp32 vanilla-PINNs with comparable accuracy. Our results show that fp64 DT-PINNs offer a superior cost-accuracy profile to fp32 vanilla-PINNs.

_posts/2021-08-14-kddpaper.markdown

@@ -65,7 +65,8 @@ <h1>
                 <a href="https://scholar.google.com/citations?user=lUmqHckAAAAJ&hl=en">Google Scholar</a> &nbsp;/&nbsp;
                 <a href="https://github.com/ramanshsharma2806">GitHub</a> &nbsp;/&nbsp;
                 <a href="https://twitter.com/ramanshsharma1">Twitter</a> &nbsp;/&nbsp;
-                <a href="https://www.linkedin.com/in/ramanshsharma">LinkedIn</a>
+                <a href="https://www.linkedin.com/in/ramanshsharma">LinkedIn</a> &nbsp;/&nbsp;
+                <a href="/docs/Ramansh_Sharma_Resume.pdf">CV</a>
               </p>
             </td>
             <td style="padding:2.5%;width:40%;max-width:40%">
@@ -126,6 +127,40 @@ <h2>Research</h2>
 
 
 
+          <tr>
+            <td style="padding:2.5%;width:25%;vertical-align:middle;min-width:120px">
+              <img src="/images/dtpinn.png" alt="project image" style="width:auto; height:auto; max-width:100%;" />
+            </td>
+            <td style="padding:2.5%;width:75%;vertical-align:middle">
+              <h3>Accelerated Training of Physics Informed Neural Networks (PINNs) using Meshless Discretizations</h3>
+              <br>
+              <strong>Ramansh Sharma</strong>, Varun Shankar
+
+              <br>
+              <em>Neural Information Processing Systems (NeurIPS)</em>, 2022
+              <br>
+              
+              <a href="/docs/dtpinn.pdf">paper</a> /
+              
+              
+              <a href="https://arxiv.org/abs/2205.09332">arxiv</a> /
+              
+              
+              <a href="https://github.com/ramanshsharma2806/dt-pinn">code</a> /
+              
+              
+              <!--  -->
+              <!--  -->
+              <p></p>
+              <p>We present a new technique for the accelerated training of physics-informed neural networks (PINNs): discretely-trained PINNs (DT-PINNs). DT-PINNs are trained by replacing exact spatial derivatives with high-order accurate numerical discretizations computed using meshless radial basis function-finite differences (RBF-FD) and applied via sparse-matrix vector multiplication. Additionally, though traditional PINNs (vanilla-PINNs) are typically stored and trained in 32-bit floating-point (fp32) on the GPU, we show that for DT-PINNs, using fp64 on the GPU leads to significantly faster training times than fp32 vanilla-PINNs with comparable accuracy. Our results show that fp64 DT-PINNs offer a superior cost-accuracy profile to fp32 vanilla-PINNs.</p>
+
+            </td>
+          </tr>
+          
+          
+          
+          
+          
           <tr>
             <td style="padding:2.5%;width:25%;vertical-align:middle;min-width:120px">
               <img src="/images/kddpaper.png" alt="project image" style="width:auto; height:auto; max-width:100%;" />
@@ -136,7 +171,7 @@ <h3>Beyond modeling: NLP Pipeline for efficient environmental policy analysis</h
               Jordi Planas, Daniel Firebanks-Quevedo, Galina Naydenova, <strong>Ramansh Sharma</strong>, Cristina Taylor, Kathleen Buckingham, Rong Fang
 
               <br>
-              <em>Fragile Earth proceedings - KDD 2021</em>, 2021
+              <em>Fragile Earth proceedings - KDD</em>, 2021
               <br>
 
               <a href="/docs/kddpaper.pdf">paper</a> /
@@ -193,6 +228,10 @@ <h2>Other notes</h2>
 
 
 
+                  
+                  
+                  
+                  
               </ul>
             </td>         
           </tr>

_posts/2022-09-15-dtpinn.markdown

@@ -5,6 +5,18 @@
 <lastmod>2021-08-15T05:29:59+05:30</lastmod>
 </url>
 <url>
+<loc>http://localhost:4000/</loc>
+<lastmod>2022-09-15T07:29:59+05:30</lastmod>
+</url>
+<url>
+<loc>http://localhost:4000/docs/Ramansh_Sharma_Resume.pdf</loc>
+<lastmod>2022-09-16T00:02:15+05:30</lastmod>
+</url>
+<url>
+<loc>http://localhost:4000/docs/dtpinn.pdf</loc>
+<lastmod>2022-09-16T00:13:21+05:30</lastmod>
+</url>
+<url>
 <loc>http://localhost:4000/docs/kddpaper.pdf</loc>
 <lastmod>2022-04-01T00:00:55+05:30</lastmod>
 </url>