July 31 post

Author: project-rheo

404.html

@@ -4,10 +4,4 @@
 permalink: /404.html
 ---
 
-<div class="text-center">
-  <h1>Whoops, this page doesn't exist.</h1>
-  <h1>Move along. (404 error)</h1>
-  <br/>
-
-  <img src="{{ 'assets/img/404-southpark.jpg' | relative_url }}" />
-</div>
+<h1 class="text-center">404</h1>

_config.yml

@@ -1,22 +1,22 @@
 # --- Basic options --- #
 
 # Name of website
-title: My website
+title: PJC 
 
 # Short description of your site
-description: A virtual proof that I'm awesome
+description: Notes
 
 # Your name to show in the footer
-author: Some Person
+author: JJ
 
 # --- List of links in the navigation bar --- #
 
 navbar-links:
-  About Me: "aboutme"
-  Resources:
-    - Beautiful Jekyll: "https://beautifuljekyll.com"
-    - Learn markdown: "https://www.markdowntutorial.com/"
-  Author's home: "https://deanattali.com"
+#    About Me: "aboutme"
+# Resources:
+#   - Beautiful Jekyll: "https://beautifuljekyll.com"
+#   - Learn markdown: "https://www.markdowntutorial.com/"
+# Author's home: "https://deanattali.com"
 
 # --- Logo --- #
 
@@ -38,11 +38,11 @@ round-avatar: true
 # Uncomment the links you want to show and add your information to each one.
 # If you don't want to show a link to an RSS feed, set rss to "false".
 social-network-links:
-  email: "someone@example.com"
-  facebook: deanattali
-  github: daattali
-  twitter: daattali
-  rss: true
+#  email: "someone@example.com"
+#  facebook: deanattali
+  github: pjchungmd
+#  twitter: daattali
+#  rss: true
 #  reddit: yourname
 #  linkedin: daattali
 #  xing: yourname
@@ -64,15 +64,15 @@ social-network-links:
 # --- General options --- #
 
 # Select which social network share links to show in posts
-share-links-active:
-  twitter: true
-  facebook: true
-  linkedin: true
-  vk: false
+#share-links-active:
+#  twitter: true
+#  facebook: true
+#  linkedin: true
+#  vk: false
 
 # How to display the link to your website in the footer
 # Remove this if you don't want a link in the footer
-url-pretty: "MyWebsite.com"
+# url-pretty: "MyWebsite.com"
 
 # Create a "tags" index page and make tags on each post clickable
 link-tags: true
@@ -168,7 +168,8 @@ date_format: "%B %-d, %Y"
 # --- You don't need to touch anything below here (but you can if you want) --- #
 
 # Output options (more information on Jekyll's site)
-timezone: "America/Vancouver"
+#timezone: "America/Vancouver"
+timezone: "America/New York"
 markdown: kramdown
 highlighter: rouge
 permalink: /:year-:month-:day-:title/

_posts/2020-02-26-flake-it-till-you-make-it.md

@@ -0,0 +1,67 @@
+---
+layout: post
+title: Extracting Symptom Data from Text
+subtitle: How can we ensure that models are reproducible?
+#cover-img: /assets/img/path.jpg
+#thumbnail-img: /assets/img/thumb.png
+#share-img: /assets/img/path.jpg
+tags: [Feinstein, NLP]
+---
+
+Extracting symptom data from text presents multiple challenges. The first is 
+identifying which conceptual entities, which may span multiple words, should be 
+labeled as a symptom, as opposed to physical exam findings, or other objective
+findings. However, prior to this, other considerations include dealing with
+very messy free text data that can contain (1) misspelled words, 
+(2) abbreviations (which may or may not be standard), (3) non-standard language
+to describe symptoms.
+
+As part of the pipeline, the text must undergo a series of transformations,
+most of which result in changes to the original text such that the it cannot be
+re-constructed from the new text. Also the transformation operators themselves
+can be subject to changes as their dependencies change over time.
+
+This leads to the problem of **reproducibility**. Perhaps a reason why being 
+able to reproduce results has lagged behind other aspects of machine learning 
+is that just being able to get things to work well for yourself is difficult
+enough, let alone getting it to work for *someone else*.
+
+To tackle this problem in the context of the project to extract symptoms data,
+let's consider how we might deal with transformed data and the transformation
+operations (which can include human programmed functions or models trained from
+data). Because we are using machine learning techniques, there are three items
+to consider: (1) data, (2) code, and (3) artifacts. While there are mature
+tools availble to version control code (e.g. Git), the process by which data
+and artifacts can and should be versioned is less straightforward.
+
+Let's consider what happens when we version control code using a tool like Git.
+Once a change has been staged, it is then commited. The commit process involves
+the programmer providing a useful comment stating what the commit was for. In
+the background, Git keeps track of the diffs between each version of the code
+and the comments provide some background information (although of varying 
+quality) as to what the change is about. Furthermore each version of the code
+is immutable.
+
+Currently the methods for version controlling data is haphazard. The most
+commonly employed technique seems to be incorporating metadata into the
+filename of the data. While a system that utilizes filename metadata might be
+useable, the main issue is that the data is mutable - files can be accidentally
+overwritten. From a data storage point of view, this can also be very
+inefficient since different files may have much in common, leading to redundant
+storage. 
+
+An issue that code versioning systems do not have to deal with is the problem
+of attributing the source of creation. The vast majority of code is written by
+humans, and therefore the source information is often found in the commit
+comment. Otherwise, if the data is forked, a clear trail exists as to where the
+code originated from. However data can come from many sources. It can be raw in
+the sense of being pulled from some outside system. It can be processed from an
+original source, and have undergone many subsequent transformations, such as
+having been processed by either code, or code with a model. 
+
+The fact that data now is often processed by models brings us back to the
+original difficulty. Models are artifacts that are the by products of both code
+and data. Without the reproducibility problem in data solved, data that is 
+transformed by models propagates this problem to the models as well (since they
+are partial offspring of data as well).
+

_posts/2020-02-28-test-markdown.md

@@ -31,13 +31,13 @@ h1,h2,h3,h4,h5,h6 {
   line-height: 1.1;
 }
 h1 {
-  font-size: 2.25rem;
+  font-size: 1.75rem;
 }
 h2 {
-  font-size: 1.875rem;
+  font-size: 1.5rem;
 }
 h3 {
-  font-size: 1.5rem;
+  font-size: 1.25rem;
 }
 h4 {
   font-size: 1.125rem;
@@ -602,15 +602,15 @@ footer .footer-custom-content {
 }
 .intro-header .page-heading h1 {
   margin-top: 0;
-  font-size: 3.125rem;
+  font-size: 1.75rem;
 }
 .intro-header .post-heading h1 {
   margin-top: 0;
-  font-size: 2.1875rem;
+  font-size: 1.5rem;
 }
 .intro-header .page-heading .page-subheading,
 .intro-header .post-heading .post-subheading {
-  font-size: 1.6875rem;
+  font-size: 1.5rem;
   line-height: 1.1;
   display: block;
   font-family: 'Open Sans', 'Helvetica Neue', Helvetica, Arial, sans-serif;

_posts/2020-07-31-symptom-extraction.md

@@ -1,6 +1,6 @@
 ---
 layout: home
-title: My website
-subtitle: This is where I will tell my friends way too much about me
+title: Mind Dump
+subtitle: Notes and Ramblings
 ---
 

assets/css/main.css

@@ -0,0 +1,3 @@
+#!/bin/sh
+
+bundle exec jekyll serve