Contains minor improvements and fix to fatal bug with EPP caused by new 1D Kullback-Leibler Divergence logic from Wayne.
1) Include all final improvements underlying our release to MathWorks of KnnFind and UMAP
2) Include EPP with some minor acceleration using KLD in 1 dimension at a time
3) Include several improvements to the old 2D gate sequence tools of AutoGated based in issues found with the OMIP 069 publication of a 40 color assay in Cytometry
1) Introduces acceleration to EPP where low KLD parameters are not explored.
2) Adds 7 new distance metrics from Connor to Jonathan’s nearest neighbor descent.
Adds faster nn descent for nearest neighbor search for UMAP for basic, supervised, template and supervised-template reduction scenarios
READIES our UMAP’s nn_descent for compilation into AutoGate production binary. UMAP reports progress to reflect nn_descent decisions and failure. If NOT compiled this reporting is supplemented when the mex fails by a popup since our open source users will want to recompile the mex source code for their environment. Currently nn_descent works for basic and supervised parameter reduction but not yet template reduction. More wandering through the random projection forest needed.
Contains improvements articulated and envisioned during grant writing
1) Full acceleration cure for Darya’s combinatorics show-stopper with matching. This makes EPP research MUCH more productive where many EPP populations match a manual population
2) Acceleration of first half of UMAP parameter reduction for basic and supervised reduction scenarios. Template reduction acceleration is next. Acceleration logic is selected when data matrix rows exceed 65,000 and columns exceed 10. Nearest neigbors must be less than 31
3) AutoResume now uses the user’s individual GoogleDrive accounts for meta data management instead of Amerige’s fragile server.
AutoResume using Google drive – QA fixes.
.prj files update for Google drive integration for AutoResume.
User’s Google drive for data store and its integration for auto resume
provides html feature that describes more details of which classes the false positive and negative mistakenly classified including background
1) add false +/- inspection graphics to MDS comparison context as well as to UMAP. Put toolbar button on sort table and a button on bottom right of histogram windows.
2) Allow user to import gates into AutoGate for false +/- and true + comparison results
1) Add support for Google Drive releases
2) Fix code synchonicity bug with UMAP where the instance variable random_state had its name and meaning change
1) add feature of finding GLOBAL highlighting colors indexed by subset name and synonym
2) fixes and improvements related to highlighting
3) fixes and improvements related to parameter reduction window
4) fixes and improvements related to statistics table sorting of gate name and gate type
continues improve manual gate revising while highlighting is active in the GatingTree and tooltips are summarizing overlap. Also some other edge case bug fixes.
Fixes and improves several UST things found when working next to Brissy at the 60th anniversary.
Continues to improve the processing of mutliple samples with templates and also makes C++ stochastic gradient descent support graphic plotting of stocastic gradient descent like the JAVA stochastic gradient descent..
improves umap supervised templates
1) In delete/switch pick lists for parameter reductions
a) Add sort cues when, duration, language, # samples & # supervisors
b) Display language in list to help debugging
2) Support html web page for all related phenograms+dissimilarity histograms for all samples related to the current supervised umap parameter reduction.
3) Avoid secondary popup windows by passing PopUp window down call chain and remembering prior text and progress bar settings
1)Decouples python implementation of UMAP from AutoGate so that open source users of FileExchange can compare the original UMAP with Connor’s to see performance differences and mapping differences.
2)When running UMAP on multiple samples make sure that the plotting of each individual sample reflect’s that sample’s data and not all of the data.
3)When processing multiple samples with UMAP allow user to create the template based on ANY combination of samples with or without probability bining and with or without supervising subsets
1)When runnning UMAP on 2 or more samples allow the user to create a supervised template on first sample
2)Support saving templates where all samples are probability binned
3)Fix bug where exception occurred when applying unsupervised template because inputData is set as a struct member.
Addresses several tricky and complicated scenarios that arise when processing multiple samples against a UMAP supervised template and some of the samples have FMO channels.
Introduces subset info import/export
Adds the C++ speedup UMAP for the MAC computer. The user must select this in the configuration panel by deselecting Show progress. This is only available on MS windows in the full MatLab environment. On the MAC this is available in both the full environment and runtime environment.
1)quicker switching between analyses
2) more options to reduce the number of files in AGI file.
3) quicker preparation of agi file
4)The ability to see disk space usage for each analysis and each kind of files in the analysis
Prevents user from restarting ParameterReduction run when using a template with more parameters than that what are available but not configured. This is now automatically adjusted without requiring user to make choices and restart the reduction.
1) Remove coefficient of separation calculation
2) Ensure closing countdown on popup windows ceases when hovering over close button and allows close to be SINGLE CLICK
3) Solve issues when applying UMAP template produced in separate GatingTree analysis
1) Support Connor’s new trans_optimize_layout.m logic to avoid false positives when applying a template whose training data differs too much from test set data.
2) Fix move_other logic so that when same embedding is head and tail both change identically. Reset the head_embedding array reference to point to the same memory as tail_embedding if move_other is true.
3) Run template from GatingTree on as many applicable samples as possible
4) Fix side-effect bugs from detaching the ParameterReduction window from the PlotEditor window. a)Ensure finish before close when closing the GatingTree window b)Ensure highlighting works
5) More new code to support running a supervised template on multiple samples from the GatingTree. 50% complete.
6) ReFactor/cleanup/normalize code for creating GatingTree gates from supervisor legend so that it is not stuck in the dbm_gui.m file that was created 11 years ago by Rachel and Noah Zimmerman..