jamesrobertlloyd gpss-research issues

jamesrobertlloyd / gpss-research

Kernel structure discovery research code - likely to be unstable

MIT License

186 stars 73 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Can I use Octave as a replacement to Matlab?

#60 ivan-marroquin opened 1 year ago
0
Discretised inputs and outputs can cause problems

#59 jamesrobertlloyd opened 10 years ago
0
Do Laplace approx on hyperparameters for increasing posterior variance

#58 duvenaud opened 10 years ago
0
Only show posterior variance

#57 jamesrobertlloyd opened 10 years ago
2
Add timestamp to all reports

#56 jamesrobertlloyd opened 10 years ago
0
Reports should cite data source

#55 jamesrobertlloyd opened 10 years ago
0
Kernels should be able to evaluate themselves

#54 jamesrobertlloyd opened 10 years ago
0
CBL parallel should check to see if big scratch has filled up

#53 jamesrobertlloyd opened 10 years ago
0
Replace infDelta

#52 jamesrobertlloyd opened 10 years ago
1
Use covPeriodicNoDC

#51 jamesrobertlloyd closed 10 years ago
0
Linear kernel translation sometimes wrong

#50 jamesrobertlloyd closed 10 years ago
1
Use spectrogram / other approaches to find candidate frequencies

#49 jamesrobertlloyd closed 10 years ago
1
Plot decomp not stable when searching over noise

#48 jamesrobertlloyd closed 10 years ago
1
Unexpected number of expansions in additive mode

#47 jamesrobertlloyd closed 10 years ago
1
Superimpose approximately periodic components to visualise the trend

#46 jamesrobertlloyd opened 10 years ago
0
Changepoints confuse discussion of monotonicity

#45 jamesrobertlloyd opened 10 years ago
0
Linear kernel restarts not great

#44 jamesrobertlloyd closed 10 years ago
1
Remove redundancy operation should occur after random restarts

#43 jamesrobertlloyd closed 10 years ago
1
Mixture of lengthscales kernel

#42 jamesrobertlloyd opened 10 years ago
0
SE lengthscale restarts

#41 jamesrobertlloyd closed 10 years ago
1
Fourier kernel not initisliased properly and numerically unstable

#40 jamesrobertlloyd closed 10 years ago
1
SE*Per translation is broken

#39 jamesrobertlloyd closed 10 years ago
1
Kernels should not have scale factors

#38 jamesrobertlloyd opened 10 years ago
0
Remove redundancy from additive kernel expressions

#37 jamesrobertlloyd closed 10 years ago
3
SE*Lin - linear standard deviation

#36 jamesrobertlloyd closed 10 years ago
1
Add an exponential function?

#35 jamesrobertlloyd closed 10 years ago
1
Translation should recognise soft changepoints and change its text accordingly

#34 jamesrobertlloyd opened 10 years ago
1
Will it ever be useful to separate the constant component of the linear kernel?

#33 jamesrobertlloyd closed 10 years ago
2
R2 can be negative when combined with changepoints

#32 jamesrobertlloyd opened 10 years ago
0
Translation should comment about the future

#31 jamesrobertlloyd closed 10 years ago
3
If we want an additive model should we always have purely additive components?

#30 jamesrobertlloyd closed 10 years ago
2
The marginal variance of covCentredPeriodic is not the output_variance

#29 jamesrobertlloyd closed 10 years ago
1
Output variances stored on log scale - make sure summary routines aware of this

#28 jamesrobertlloyd closed 10 years ago
0
cblparallel should sanitise its code input

#27 jamesrobertlloyd opened 10 years ago
0
XKCDify output

#26 jamesrobertlloyd opened 10 years ago
0
Sum, Product, Changepoint should be operators

#25 jamesrobertlloyd closed 10 years ago
1
Allow min period heuristic type to be an experiment parameter

#24 jamesrobertlloyd closed 10 years ago
1
More data

#23 jamesrobertlloyd opened 10 years ago
3
Try the log transform of some data

#22 jamesrobertlloyd opened 10 years ago
0
English language kernel descriptions

#21 jamesrobertlloyd closed 10 years ago
1
Hessian should be computed on full data when subsetting?

#20 jamesrobertlloyd closed 10 years ago
1
Limit the number of decomposition components plotted

#19 jamesrobertlloyd closed 10 years ago
1
Should the search always use the best kernel found so far for expansion?

#18 jamesrobertlloyd closed 10 years ago
1
Mask kernels should allow None dimensions

#17 jamesrobertlloyd closed 10 years ago
1
Solar takes a surprisingly long time to expand at lower levels

#16 jamesrobertlloyd closed 10 years ago
2
Reintroduce the laplace approx

#15 jamesrobertlloyd closed 10 years ago
1
Make suitable for multi-d again

#14 jamesrobertlloyd opened 10 years ago
0
Combine anticorrelated components

#13 jamesrobertlloyd closed 10 years ago
1
Expand all subkernel expressions

#12 jamesrobertlloyd closed 10 years ago
1
Check that python is using the random seed correctly

#11 jamesrobertlloyd closed 10 years ago
1