Bummer! You're not a
Stitcher Premium subscriber yet.
Learn More
Start Free Trial
$4.99/Month after free trial
HELP

Show Info

Episode Info

Episode Info: ## Episode - Hypers future & meta-learning ** We're always removing hypers. DL removed feature-engineering ` - Model selection ** Unsupervised? K-means Clustering => DL ** Linear? Linear regression, logistic regression ** Simple? Naive Bayes, Decision Tree (Random Forest, Gradient Boosting) ** Little data? Boosting ** Lots of data, complex situation? Deep learning - Network ** Layer arch ** Vision? CNN ** Time? LSTM ** Other? MLP ** Trading LSTM => CNN decision ** Layer size design (funnel, etc) ** Face pics ** From BTC episode ** Don't know? Layers=1, Neurons=mean(inputs, output) https://stats.stackexchange.com/a/1097/107199 - Activations / nonlinearity https://towardsdatascience.com/activation-functions-neural-networks-1cbd9f8d91d6 ** Output ** Sigmoid = predict probability of output, usually at output ** Softmax = multi-class ** Nothing = regression ** Relu family (Leaky Relu, Elu, Selu, ...) = vanishing gradient (gradient is constant), performance, usually better ** Tanh = classification between two classes, mean 0 importantRead more ยป

Discover more stories like this.

Like Stitcher On Facebook

EMBED

Episode Options

Listen Whenever

Similar Episodes

Related Episodes