Charles Martin

Similar to the python powerlaw package ? if not, what the plmle functon in the new toolbox called ?

Please remove print statement on line 341 of powerlaw.py and place this into a python warning or info log so it can supressed when running a very large number of...

divide by 0 warnings

1

let me suggest using ` (Theoretical_CDF * (1 - Theoretical_CDF)+0.0000001) ` on line 700 of

Can not compile on MacOS High Sierra, using Python 3.10

1

Can not compile on MacOS High Sierra, using Python 3.10 cpp --version Apple LLVM version 9.0.0 (clang-900.0.38) Target: x86_64-apple-darwin17.7.0 **Here's the error message:** building 'ad3.factor_graph' extension clang -Wno-unused-result -Wsign-compare -Wunreachable-code...

add exponential decay for weight gradients

grid search parametets

2

It is unclear what makes a good initial starting point for the SVM We can grid search the parameters to find out try testing different values of sigma, learning_rate, momentum,...

omniglot test does not use rbm features

2

We use the RBM to generate features for the SVM or Logistic Regression so we need to transform X_train to features (F_train): example: rbm = BernoulliRBM() rbm= rbm.fit(X_train) F_train =...

pytest replicates data dir

1

data dir should not be replicated I use a symbolic link data -> ../data/

test_omniglot does not binarize the data

2

See EMF_RBM_Energy_Densities_Ominglot-200 Data should be binary to run in a Bernoulli RBM You can see this by plotting the data