From fcad2d81cc426cbde379ff1f2aa3eb5914a1110b Mon Sep 17 00:00:00 2001 From: wassname Date: Sun, 16 Feb 2020 16:49:25 +0800 Subject: [PATCH] readme --- readme.md | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/readme.md b/readme.md index cf4903f..6508ba4 100644 --- a/readme.md +++ b/readme.md @@ -28,11 +28,13 @@ I've also made lots of tweaks for flexibility and stability and [replicated the - [Example NP](#example-np) - [Example ANP outputs (sequential)](#example-anp-outputs-sequential) - [Example ANP-RNN outputs](#example-anp-rnn-outputs) + - [Example of ANP-RNN with MCDropout](#example-of-anp-rnn-with-mcdropout) - [Replicating DeepMind's tensorflow ANP behaviour](#replicating-deepminds-tensorflow-anp-behaviour) - [Using Monte Carlo Dropout](#using-monte-carlo-dropout) - [Usage](#usage) - [Smartmeter Data](#smartmeter-data) - [Code](#code) + - [ANP-RNN diagram](#anp-rnn-diagram) - [See also:](#see-also) @@ -90,11 +92,10 @@ This has a better calibrated uncertainty and a better fit ![](docs/anp-rnn_4.png) -A Diagram of ANP-RNN's - -![](docs/np_lstm.jpeg) +### Example of ANP-RNN with MCDropout +![](docs/anp-rnn-mcdropout.png) ## Replicating DeepMind's tensorflow ANP behaviour @@ -177,6 +178,10 @@ Changes for stability: - check and skip nonfinite values because for extreme inputs we can still get nan's. Also gradient clipping - use pytorch lightning for early stopping, hyperparam opt, and reduce learning rate on plateau +## ANP-RNN diagram + +![](docs/anp-rnn-mcdropout.png) + ## See also: A list of projects I used as reference or modified to make this one: