Woody Kim

Results 24 issues of Woody Kim

need to stop train when nlp networks (seq2seq, word2vec, doc2vec) reaches certain goal to prevent overfitting and save cpu powers

for now h5 files feeds in sequence, but we need to shuffle it to improve model accuracy

from : /api/v1/type/common/target/nninfo/ to : /api/v1/type/common/target/nninfo/nn00016// reason is you cannot pass search condition info with json when you call rest api with get method on real browser(it works fine with...

1. eval common function that manage eval result and handle db transaction 2. actual eval functions for each neural networks 3. on prediction time we need to cache model and...