!self reflection

2019-12-12

work notes

2019-12-12 - summary the experiences from HUS

2018-12-02

  • chat bot
  • Image caption

2018-

  • latest two years from 2017

  • check the review number 10W+

  • the size of the model

  • dell user review data

  • config, reviews

2018-06-01

  • query=attributes traing using MLP
    • softmax
    • sigmoid for the binary classification

2018-05-31

  • read the paper - "predicting latent structured intents from shopping queries" - MLP

    • RNN
    • LSTM

  • claw the review data - "small" subsets

  • run basic word2vec on TF

2018-05-30

  • word embedding

    • data normalization

    • lemmatization and word stem ?

    • document vectorization

    • count vectorizer and TF-IDF vectorizer

    • word2vec <- could care about the order (word context) -> semantic
      -> we get the similar vectors for the words

2018-05-29

meetup with Dr. Wang on 28/05/2018

  • about tensorflow [1]:

    • Setup the env about python and tensorflow

    • now a simple tensorflow could be run in locally -> see the github [4]

    • plan (todo):

      • run a basic mode like linear regression in tensorflow
      • then run word2vec in tensorflow, or try to run it on server

  • about word embedding [2][3]:

    • the basic applied machine learning knowledge: like loss functions, bag of words, features, bag of vectors

    • if there is something wrong, if we could know the principle/theory, we could know the reason and correct it quickly

    • plan (todo):

    • know more about ML, especially deep learning (like word embedding part) based on the reference 2 and 3

  • Others

  • the github for the code
  • the fixed meetting time

2018-05-24

  • setup the env on python and tensorflow on mac
  • TODO:
    • setup the env on windows

virtualenvwrapper -> export WORKON_HOME=$HOME/.env

  • For some project

    cd dev.dplearning

  • create an env

    mkvirtualenv tfenv -> including the Python executable files, and a copy of the pip lib

  • use the env

    workon tfenv -> using the virtual env
    pip install *

  • deactivate

    deactivate -> deactivate the env
    rmvirtualenv venv

pip freeze > requirements.txt

pip install -r requirements.txt

reference