Difference between revisions of "User:TChen"

From REU@MU
Jump to: navigation, search
(Work Log)
Line 138: Line 138:
 
'''''Monday 7/26'''''
 
'''''Monday 7/26'''''
 
*Still working on Kriging interpolation
 
*Still working on Kriging interpolation
 +
*Encountering difficulty in aligning metrics between ATL03 and ATL07
 +
'''''Tuesday 7/27'''''
 +
*Met with Satish to discuss upcoming plan for the week
 +
*Looked into LSTM for time-series predictions
 +
*Successfully replicated ATL07 classification with 97% accuracy (but at the same granularity as the ATL07 data)
 +
'''''Wednesday 7/28'''''
 +
*Mainly reading LSTM papers
 +
'''''Thursday 7/29'''''
 +
*Returned to Kriging vs learning comparison

Revision as of 23:32, 29 July 2021

About Me

Hello! I'm Theresa Chen, student at Carleton College, class of '22. I am a double major in environmental studies and computer science.

Work Log

Week 1

Working part-time due to still being in school
Tuesday 6/1

  • Attend REU orientation
  • Attend mini data science boot camp

Wednesday 6/2

  • Continue mini data science boot camp

Thursday 6/3

  • Attend Good Research Practices Talk by Dr. Brylow

Week 2

Tuesday 6/8

  • Meet with Dr. Puri, discuss overall research goals
  • Begin literature review

Wednesday 6/9

  • Responsible Conduct of Research (RCR) training
  • Attend REU event on technical writing

Thursday 6/10

  • Continue lit review
  • Familiarize self with different ICESAT-2 data products
  • Watch some tutorials from ICESAT-2 hackweek

Friday 6/11

  • Lit review- looking at articles related to areas of interest in the arctic
  • Meet with Dr. Puri
  • Narrow down interest to ATL03 files, start pre-processing data

Week 3

Monday 6/14

  • Continue lit review
  • Explore captoolkit, used to process ATL06 open altimetry data
  • Try to reduce HDF5 file to only variables selected for project

Tuesday 6/15

  • Visualize raw photon height data from selected study area in the arctic
  • Experiment with data visualizations
  • Review past paper from old REU participant Erin Doyle
  • Meet with Dr. Puri, decide to expand variable selection

Wednesday 6/16

  • Working on processing HDF5 to csv
  • Went to REU Professional Event
  • Created REU Wiki pages

Thursday 6/17

  • Double checking data sets by visualizing photon heights to see if elevation makes sense
  • Obtaining datasets for several areas near the arctic circle
  • Beginning machine learning tutorials using "keras" package

Friday 6/18

  • Working on creating neural network
  • Mostly debugging tensorflow installation
  • Experimenting with hyperparameters like epochs and batch size to reduce loss and increase accuracy

Week 4

Monday 6/21

  • Predicted results from NN still have very low accuracy, still trying to fix

Tuesday 6/22

  • Fixing uneven output types for NN model
  • Fixing low accuracy for NN model
  • Met with Satish to discuss expectations for the coming week

Wednesday 6/23

  • Finally got working results for NN model
  • Working to visualize and compare results for height prediction
  • Learning how to use geopandas and geoplot

Thursday 6/24

  • Debugging geoplot, issues with conda installation
  • Working on data visualizations
  • Started using cartopy

Friday 6/25

  • Lit review on lead classification
  • Finalized some data visualizations

Week 5

Monday 6/28

  • Attended RCR on presentations
  • Worked on slides for presentation on Wednesday
  • Read a paper on calculating sea ice thickness from ATL10 freeboard estimates

Tuesday 6/29

  • Read a paper on lead classification
  • Met with Satish to discuss verification of results
  • Practiced for Wednesday presentation

Wednesday 6/30

  • Attended research presentations and gave presentation in the morning
  • Compared results to OpenAltimetry results
  • Read up on CRYOSAT and Sentinel-1

Thursday 7/1

  • Read up on ATL10 Freeboard algorithm
  • Looked at alternative machine learning classification techniques for lead detection

Friday 7/2

  • Continued researching lead detection
  • Set up Marquette VPN so that larger area in arctic circle could be analyzed
  • Read literature on features and areas of interest in Arctic Circle

Week 6

Monday 7/5

  • Lit review of CRYOSAT-2 lead detection algorithms
  • Review of decision trees

Tuesday 7/6

  • Attended training on ethical computing
  • Code reorganization so that it will be scalable to multiple HDF5 files
  • Meeting with Satish to discuss building a lead classifier

Wednesday 7/7

  • Attended training on making research posters
  • Started coding neural network for classification

Thursday 7/8

  • Encountering difficulties with classification due to photon aggregation
  • Reading up on how much granularity is possible

Friday 7/9

  • Working on processing ATL07 files
  • Exploring different types of classifiers

Week 7

Monday 7/12

  • Fixing data issue with ATL07
  • Familiarizing self with Beaufort sea area

Tuesday 7/13

  • Met with Satish to discuss data issue within ATL07 and ATL03
  • Read up on CNN
  • Read up on Gaussian Width calculations

Wednesday 7/14

  • Recalculating Gaussian Width is not possible
  • Attended RCR training in the morning

Thursday 7/15

  • Using icepyx API to download Beaufort sea coordinates

Friday 7/16

  • Looking at time series analyses as a way to analyze data
  • Trying to map out four year time series of Beaufort Sea

Week 8

Monday 7/19

  • Continuing to struggle with icepyx (doesn't accept multipolygon shapefiles?)
  • Continuing to struggle with data wrangling in order to directly compare ATL07 and ATL03
  • Trying to fix data issue, not succeeding, is there any way to count photons in segment?

Tuesday 7/20

  • Finally imported data set to remote server!

Wednesday 7/22

  • Read paper on applying persistent homology as an analytical approach to data

Thursday 7/23

  • Kriging interpolation

Friday 7/24

  • Reconstructing ATL07 classifier

Week 9

Monday 7/26

  • Still working on Kriging interpolation
  • Encountering difficulty in aligning metrics between ATL03 and ATL07

Tuesday 7/27

  • Met with Satish to discuss upcoming plan for the week
  • Looked into LSTM for time-series predictions
  • Successfully replicated ATL07 classification with 97% accuracy (but at the same granularity as the ATL07 data)

Wednesday 7/28

  • Mainly reading LSTM papers

Thursday 7/29

  • Returned to Kriging vs learning comparison