MLCC #2
0 likes123 views
This document discusses machine learning concepts including mini-batch stochastic gradient descent, IID datasets with stationary distributions, good features that show up frequently and have clear definitions without outliers, and an example of using Keras to predict traditional Chinese medicine prices from origin data by building a dataset, running a regression model, and saving the model.
MLCC #2
- 1. MLCC #2 Ofa
- 2. SGD • batch size • SGD / mini-batch SGD / full batch
- 3. Datasets • IID • stationary distribution • always pull from the same distribution
- 4. Feature • Good feature • show up with reasonable frequencies • clear def & meaning //static def • no magic value • Feature engineering • scaling • outliers • one-hot encoding • know your data
- 5. Example - using keras to predict TCM price from origin Step1. import libraries
- 6. Example - using keras to predict TCM price from origin Step2. build dataset data from origin data from store standard data training
- 7. Example - using keras to predict TCM price from origin Step3. run regression Step4. save model and weights