upload

MEME/index.html

@@ -225,6 +225,28 @@ <h1 class="title is-2" style="text-align: center; padding-bottom: 10px;">Multipl
           </div>
         </div>
 
+        <div class="column" style="text-align: center;">
+          <img src="resources/model.png" alt="Image description" width="60%">
+      </div>
+        <br>
+        <br>
+        <h2 class="subtitle is-4" style="text-align: center;">Why Multiple Embeddings</h2>
+        <div class="columns is-centered">
+          <div class="column is-full" style="text-align: center;">
+            <br>
+            <br>
+            <div class="columns is-centered has-text-centered">
+              <div class="column is-two-thirds">
+              <div class="content has-text-justified interpolation-panel">
+                  <p style="text-align: center;font-size: 18px"> Our research demonstrates that multimodal embeddings offer superior performance, partly due to their ability to handle shorter context lengths a current bottleneck of the BERT architecture. We initially hypothesized that the use of multiple embeddings allows for comprehensive representation across diverse biological and temporal scales and different domains. By employing separate embeddings for molecular scale data (such as lab measurements and medications), system-level observations (including vital signs and diagnostic decisions), and external influences (like patient arrival times), we can capture unique aspects of patient attributes. This method provides a holistic understanding and enhances the effectiveness of patient care management, serving as our rationale for representing these concepts distinctly.</p>
+              </div>
+            </div>
+          </div>
+          <br>
+
+      </div>
+    </div>
+
 
 
   <section class="section" id="BibTeX">