Monday, February 6, 2012

Winter Meeting: Monday Morning - DIY Tech, PER, Physics in Animation

After a very welcoming "First Timers" breakfast, I made my way to the "DIY Technology for the Physics Classroom" session, where I saw three talks.

General observations (going into my thinking process about my talk tomorrow):
  • There are kind of a lot of folks at each talk
  • 10 minutes, especially compared to the usual 95 minutes that I get in class, is not a lot
  • I hope that the contrast for the Osmos clips is good in these bright rooms
  • It would be nice to include some expanded references to the Osmos talk for reference.  I'll see if I have time today to get that done!
 IPAL: In-class Polling for All Learners
  • Bill has developed some open-source polling software that can be used with any web-enabled device - no clickers required
  • The Moodle or stand-alone modules can import ComPADRE or other established question sets, so you have something to start with at least!
  • Bill Junkins is happy to help you out with this:

 Using Your Classroom Projector to Demonstrate Some Properties of Light:
  • Wow: DLP projector technology seems to be overly complex, but very neat: strobe light + rotating color wheel + tons of tiny individually-aimable mirrors!
  • The LCD projector's dichroic combiner cube looks like a good total internal reflection demo/application for class
  • Using holographic diffraction grating glasses (cheap!), which are 500 line/mm diffraction gratings, looking at vertical lines through the projector is a neat demo n color mixing, and also showed that we didn't have an incandescent bulb in the projector
  • Using a polarizer, you can show that the projector's green light is polarized perpendicularly to the red and blue!
  • For more info:
 Build your own electric field demonstrator
  • This is supposed to be a simpler demo than the "grass seeds in mineral oil" demo
  • This demo uses lettuce seeds instead (more arrow-like) and vegetable oil instead, with a VdG to set it up, along with a hangar, styrofoam cups (to stand-off the hanger between the vdG and petri dish.  Put paper under dish for camera mounted above, use fluorescent tube to bleed off charge
  • It should spark between the wires, and the seeds will move quite a bit
  • He showed videos of monopole source, dipole, parallel plates, absence of field inside a conductor (my favorite)
  • For more info:
After this, I went to the "PER: Investigating Classroom Strategies" session:

Getting the Word Out: Effective Communication of PER Study Results
  • This is about not original research, but best communicating the results to others; she's a science journalist as well, so uses lessons from that field
  • The fundamental disconnect between being aware of data and an actual change in behavior is really analogous to the whole "book learning" issues for students and, apparently, for teachers taking PER results seriously - many teachers know about it, but still continue to lecture
  • I want to DL this one to share.
  • The information-laden approach that we try to use to advocate for research-based instruction is just as big a problem as trying to do it with students
  • "The deficit model assumes that the public are empty vessels waited to be filled with science knowledge, upon which they will rationally act" (OK, I typed it quickly - maybe not a quote)
  • Unfortunately, initial emotional appeals, etc. with some data as support are more effective than a data-driven approach
  • Be upfront about the challenges of implementation - many try reformed strategies, only to drop them after a term or so
  • Contact:
The Challenges of Assessing Teaching Effectiveness: Strategies for PER to Influence Practice
  • Importance is for quality assurance (from the institutional side) and for personal teaching reflection and modification
  • Aligning the assessment instruments for those purposes (and aligning them to PER) is important
  • This alignment isn't usually present, though: lots of reliance on student evaluations and other poor measures by institutions. External measures, like the US News rankings, are super-terrible
  • The heavy use of student evaluations makes folks nervous to change things for fear of lower evaluations (even if the measures of learning increase!)
  • We need more and better-publicized research-based assessments
  • PER is good as the lesson/course level: was the lesson/course successful?  It doesn't have as much for program-level assessments, but there's starting to be focus on this at a political level, so we need to catch up
  • Measuring progression of reasoning skills and conceptual understanding of specific concepts across a program are easier; we don't have much for problem-solving skills yet - what would that look like?
  • Contact:
Variance and Variables: The Analysis of Pre-test Results from Thousands of Students
  • It must be nice to have a gigantic data set (like several introductory courses in a large college)!
  • We're comparing pre- and post-test data
  • First question: are they normally distributed?  Normal quantile plot
  • ...or binomially distributed? (as if they were draw from a population of students giving random correct/incorrect answers)  Nice presentation - there was probably the biggest binomial-gaussian distribution comparison graphical analysis audience laugh ever heard.
  • Amount of prior instruction: ANOVA analysis showing differences between no instruction, some instruction, and all instruction; not much difference in many cases
  • Open question: why are some questions binomially distributed and others not?
  • Contact:
Designing Research-based Instruction in a Large Lecture Course Without Recitations
  • Intro calculus-based college course: 4 lectures, no recitations or TAs, lab not run by teacher
  • HW assignments online, most tests multiple-choice
  • In such a traditional, inflexible environment, how can we use PER?
  • Prelectures/checkpoints (multimedia, animations, etc.): works better than text reading, students like them
  • LON-CAPA and some turned-in work (mostly on form)
  • You-tube videos of worked problems - takes that out of classroom
  • <10 minutes of actual lecture
  • 3-8 daily voting questions, using colored cards instead of clickers
  • Important strategies: leaving the stage, answering questions, monitoring student discussions
  • FMCE gains weren't super-duper, but the checkpoints and cards helped somewhat
  • Some research cited on "high structure" classrooms, and how that helps at-risk students
  • Nice distinction: urban, rural, and... frontier - he does teach in North Dakota.  Nice graphic on >50% of the counties in ND which have <= 6 people per square mile!
Learning Integration in Physics Using Debate Problems and Multimodal Communication
  • Looking at center-of-mass determination
  • Conceptual intro to continuous distributions: here's the discrete definition, which is the continuous definition? (poll question)
  • After that's done, how do you determine if they really have it?  A couple of diagrammatic methods shown, ask them to derive the integral def'n from the discrete (hard), ask them to apply the integral equation (not the same as understanding what it means), have them use multiple representations (diagrams, etc.) - best!
  • 1. Group discussion; 2. Groups present to each other; 3. Recorded presentations shared between sections
  • They assessed individual student presentations with Smartpens, so you have to paper and the student's narration
  • Another example question: I for rod; determine integral expression for rod around parallel axis
  • Result - lots of students think that 'd' in an integral means 'change', and maybe this is the first static integral they've ever seen! 
The next place I went was to a talk given by Ron Henderson (physics PhD, Princeton, formerley worked at Cal Tech) on how waves are encoded in animation.  Ron's a DreamWorks animator.
  • 22,000 Intel Xeon cores, 120 TB or storage, 120,000 frames, >1 billion files for a single DreamWorks movie - yikes.
  • The animation rig for a single character in one of these movies can have up to 10,000 controls - lots of strings for a puppet to have
  • 200-800,000 hairs for a rendered cat, like Puss in Boots.  ...then you have to make the hair go around the belt, etc.
  • Goose feathers: 'grown' via algorithm along equipotential lines, in order to minimize collisions between them (just like real goose feathers, approximately). So there.
  • Volumetric modeling of clouds for high-res 3D models, like in the Jack and the Beanstalk bit
  • 34% Bachelor's, 40% Masters, 20% PhD among R and D engineering staff; degrees most common in CS, Engineering, graphics, match, physics (4%), in that order
  • Premise: non-physical motion is distracting to audiences.  Why don't they hate Star Wars then? :)  It all really depends on knowing what your audience knows and doesn't know!
  • A good way to plan out a complex endeavor: they look at each movie and ask themselves: "what do we want to go after that we haven't gone after before?"  You can't reinvent every wheel every time.  Pick your battles.
  • Breakdown of man-hours for effects for Puss in Boot (in order): clouds, dust, destruction, water, splashes, tornado, beanstalk
  • Their destruction model includes: explosive, primary debris (moment of inertia, spin simulation here!), secondary debris, dust source, fluid simulation, and visual development; fracture analysis, particle dynamics, rigid body dynamics, and fluid simulation are the primary models for destruction animation
  • Their model selection is artist-driven, so multiple approaches/representations are very important for them
  • There's a nice explicit balance between physical law (simulation) and computability (using simplified models).  This is a good thing to emphasize with students: we understand the approximations and simplifications that we use (and why!), but we can usually do a better job of getting students involved in that process!
  • Incompressible Euler equations sued for tornado simulation; apparently a common model for fluids in animation.  They're controlling a buoyancy term, a dissipation term, and a third term.  
  • The complex behavior of the tornado comes from solving the equation for lots of little boxes - the coupling between the boxes creates the complexity.
  • They let the artists control divergence terms to make it look how they want it.  They add obstacles to respond to the artist requests to "make it more turbulent"
  • They warp the physical simulation onto a controlled skeleton to compromise between the desired physical and un-physical parts of the animations
  • Great question from the crowd - "there's lots of physics going into making the animations look great - is there a lot of biology going into the beanstalk animation (and others)?" Answer: "no."  There are skeletons, etc. but skin and such are algorithmic at this point (designed to "look right", rather than physical simulations).
  • This was pretty awesome.  The 45 minutes flew by - I could stand another couple of hours of this! to In-N-Out Burger!

    No comments:

    Post a Comment