7:00 – 9:00, 10:00-4:30 IRAD

• Something for my Visible from Space collection?
• Saw this in the letters from CACM this month:
  

  Causal Connections for Predictive AI

  As part of the “ACM Panels in Print” section “Artificial Intelligence” (Feb. 2017), panelist David Blei said he believes computer science needs to identify the causal connections between data components, concluding that artificial intelligence, with its predictive capabilities, will be enhanced through causal inference. For example, the first step toward using AI in database systems is to analyze and create a data map of the complexity of the causal interconnections between the data components in a problem space. A data map connects a data component to other data components through causal interrelationships. A data map can be created through qualitative analysis of data collected for a particular problem. The qualitative analysis could then take the form of “thematic analysis,” (Braun, V. and Clarke, V. 2006) using systems diagramming to gain greater insight into the data. At that point it might be advantageous to start applying AI directly to the data.

  Researching the complexity of database systems, I have thus created such a data map, which is now ready to move to the next stage of automation where predictive analytics can help improve management of database systems. Using an analogy of the CODEX, or Control of Data Expediently, my research into causal connections has identified a potential role for AI in automating continuously changing best practice, thus representing an agile approach to deciphering the complexity of interconnections and promising to help create an autonomous way to deliver best practice in database management.

  Victoria Holt, Bath, U.K.

• I looked at hew blog, research page and twitter account. Can’t find anything about mapping. Even looking at the thematic analysis and citing articles, I can’t find much about maps or mapping. Odd
• Getting back to drawing the database. But first, updating my IDE, it seems…
• Broke out DimensionInfo to TensorDimensionInfo since I need the mappingLow and mappingHigh to draw the borders
• Changed the canvas drawBG to use StorageAndRetreival, since I need to know the projection I’m drawing
• Killed JavaFX for a while. Had to reboot
• More RFI work
• Ok, got the background scaled correctly and drawing the border. Had to deal with more hard-coded values. SAD! Anyway, here’s all the parts working: 

6:30 – 6:00 ASRC Research, proposals and job stuff

• Redid my resume this morning. The word doc went fine, but when I went to generate the html versions, I had a momentary lapse of good sense and had Word output the pages.  Horrible mistake. I had to go and use the very fine HTML Cleaner website and put everything back together again. So 15 minutes of adding a paragraph and a picture turned into over an hour of work. Sigh.
• If I have time, the next step today is to get the StorageAndRetrieval code drawing on the canvas. Started setting this up. The actual cell coloring could be done with another evaluateTensor method? Maybe later. Get the basics working first.
• In a significant miracle, my PhD work is now my job, at least until the end of the year. Sprinkled with some proposal writing.
• Spent most of the day working on the NNSA RFI

7:00 – 6:30 ASRC Research

• Probably not going to get the StorageAndRetreival class done and integrated by Sept 1, so I’ll need to work out the paper based on what I have. But let’s see how far we get…
• I think I’m going to start simply by placing the agent type in the IR cell and color accordingly. There will be three states, NULL (white) , EXPLORE (red) and EXPLOIT (green)
• Fixing some things along the way. Stage dimension initial size was hard coded. Oops.
• Adding a clear() to LabeledTensor and StorageAndRetreival
• Working out how to recurse though a set of arrays. Putting here for later reference:

  class Recurse{
      int linecount = 0;
  
      boolean fillMat(int[] curIndices, int[] maxIndices, int curIndex){
          if(curIndex > curIndices.length-1){
              //System.out.printf("curIndex(%d) > %d\n", curIndex, curIndices.length-1);
              return true;
          }
          for(int i = 0; i < maxIndices[curIndex]; ++i) {
              curIndices[curIndex] = i;
              boolean done = fillMat(curIndices, maxIndices, curIndex+1);
              if(done) {
                  linecount++;
                  System.out.println(linecount+" [" + curIndex + "]: " + Arrays.toString(curIndices) + " = " + done);
              }
          }
          return false;
      }
  
      public static void main(String[] args){
          int[] initial = {0, 0, 0};
          int[] max = {3, 3, 3};
          System.out.println("Recurse!");
          Recurse r = new Recurse();
          r.fillMat(initial, max, 0);
      }
  }

• Here’s the final version, with an Interface callback

  public boolean fillTensor(int[] curIndices, int[] maxIndices, int curIndex, TensorCellEvaluator tce){
      if(curIndex > curIndices.length-1){
          // if we get here, we're done
          return true;
      }
      for(int i = 0; i < maxIndices[curIndex]; ++i) {
          curIndices[curIndex] = i;
          boolean done = fillTensor(curIndices, maxIndices, curIndex+1, tce);
          if(done) {
              tce.evaluate(curIndices, lt);
          }
      }
      // if we get here, we're not none, keep going
      return false;
  }

• Here’s the call with the implemented interface. It doesn’t do anything yet, but the pieces all work

  stoRet.fillTensor(indicies, maxIndicies, 0, new TensorCellEvaluator() {
      @Override
      public String evaluate(int[] indicies, LabeledTensor lt) {
          System.out.println(Arrays.toString(indicies));
          return Arrays.toString(indicies);
      }
  });

• Interview with USDA team. Not a fit at all

Another weekend of scrambling to get everything done in time. Plus a good bike ride 🙂

7:00 – 5:00 Research (ASRC IRAD)

• Started up the computer this morning and it freaked out. After a restart, it rebuilt all the program indices and many of my configs were gone. I’m guessing that it was a bad disk sector. Scary, since this is a pretty new box
• Adding in the Information Retrieval part that will mediate interactions between agents
• And now the changes to JavaUtils2 aren’t being recognized. Deleted reference and reimported. And not the project is rebuilding super slow????? This is just not my morning.
• And now IntelliJ is frozen. Restarted. The error is still there, but it compiles no problem. Had to invalidate caches.
• Started StorageAndRetreival class and have a rough framing running
• Long discussion with the interns about using mapping for the HR space
• Wrote up some more paragraphs on normative mapping for paper justification
• Fuzzy thought for today. The map is actually a set of pointers to the data used to produce it. For example, a discussion that produces normative poles has a history of collapse that the normative poles could reference back to.

7:00 – 5:00 ASRC IRAD

• Got Understanding Ignorance, which has a section on Ignorance as a Place. More on that later
• Since I’m in the middle of LabledTensor, I’m going to add the ability to string together multiple ArrayLists to accommodate Long indexing
  • Created class LongArrayList, based on part of List. Can’t implement the interface, since the parameters are int, rather than long. Done! Also fixed a few formatting errors in the toString().

6:45 – 3:00 ASRC IRAD

• A Physicist Who Models ISIS and the Alt-Right
• Working on adding the labels to LabeledTensor. Making an inner class that handles all the unique pieces of the dimension. It will also handle the checks, etc. Mostly done, but I need to make sure it handles axis of different size.
  • Well drat: “Our operations teams are currently working to mitigate a Distributed Denial of Service (DDoS) attack on our networks. ” I can’t get to SVN. Yay! Got it!
  • It just struck me that if I make this work with MapReduce, I can have Geomesa in something like 500 lines of code. I think I’ll call it GeoPlain
  • Got different length axis working
• Important thing to think about with respect to maps: I find word clouds strange b/c they arrange summaries of text data in a way we would never handle summaries of numeric data. 
• Data-driven Advice for Applying Machine Learning to Bioinformatics Problems
• 10:00 – 12:00 Song Chen’s  Fraud Detection in Healthcare dissertation defense.
  • Fiedler Vector
  • What about spikes, such as a particularly bad flu or natural disaster? Is there a sense of normal at a particular time?
  • bias in synthetic data?
  • Community detection
  • There are some good public datasets for testing healthcare. Heritage Foundation Prize dataset is used most in this work
  • cosine similarity by procedure codes
  • Is community pruning  a generally effective optimization technique?
  • What if community membership changes over time? Fraudulent doctors who move to a different state, for example?
  • Parallel Heuristics for Scalable Community Detection

Eclipsing!

A multidimensional framework for studying social predation strategies

Individual behaviour in isolation does not match individual behaviour in the social context!

7:00 – 8:00 Research

• Assembling my C&C notes here. Done. Need to review, strip the raw notes out of my paper and integrate.
• Oxford Internet Institute – Call for Papers: Policy and Internet Special Issue on Reframing ‘Fake News’: Architectures, Influence, and Automation Due Oct 31. Length of 6k – 8k words

8:30 – 6:00 BRI

• Remembered my bike shoes, but it may rain all day
• Wrote up a note to T about burning PhD hours for the Oct1 CHIIR paper deadline
• Need to write up the dev plan for GeoMesa into the EIP pattern (include a strawman use case?)
• Back to learning Spring
• And here is a way to populate a nested bean:

  <bean id="countryParent" abstract="true" class="org.springframework.chapter2.pojos.EmployeeImpl"><!-- this just gets one abstract class implementation -->
      <property name="country" value="USA"/>
  </bean>
  
  <bean id="employeeServiceBean"
        class="org.springframework.chapter2.service.EmployeeServiceImpl">
      <property name="employeeList">
          <list>
              <bean parent="countryParent" class="org.springframework.chapter2.pojos.EmployeeImpl">
                  <property name="employeeAge" value="10"/>
                  <property name="employeeName" value="Ten Tin"/>
                  <property name="employeeId" value="1"/>
              </bean>
              <bean parent="countryParent" class="org.springframework.chapter2.pojos.EmployeeImpl">
                  <property name="employeeAge" value="11"/>
                  <property name="employeeName" value="Eleven Even"/>
                  <property name="employeeId" value="2"/>
              </bean>
              <bean parent="countryParent" class="org.springframework.chapter2.pojos.EmployeeImpl">
                  <property name="employeeAge" value="12"/>
                  <property name="employeeName" value="Twelve Elve"/>
                  <property name="employeeId" value="3"/>
              </bean>
          </list>
      </property>
  </bean>

• That sends the instanced objects to setEmployeeList that has to take a list of objects. To get it typed, I did this:

  @Override
  public void setEmployeeList(List<Object> employeeList) {
      for(Object o : employeeList){
          Employee e = (Employee)o;
          setEmployee(e);
      }
  }

• A typed employee is added as follows:

  @Override
  public void setEmployee(Employee employee) {
      employees.add(employee);
  
  }

• IntelliJ will extract an interface and rename your class. How cool!
• Printed, signed and scanned my credit card agreement.

8:30 – 3:30  BRI

• Got in at 10:30 last night and am here on time. Kind of surprised. I’ll need to leave a bit early so that I can get shopping and lawn triaged. It rained a LOT while I was gone.
• getting caught up, updating IDE, etc. It looks like I’m still working on GeoMesaIngest.
• Since I still need to come up to speed on Spring, and I don’t have tasking, I think I’ll start there
• Downloaded book
• installed lamp-server^ and php. Verified that everything is running
• Got an end-to-end Spring example running with no mysterious incantations. Nice. Now I want to understand testing, but I have to go through a few chapters first…