Possible bug in "lesson3-rossman.ipynb"

danaludwig · July 16, 2018, 4:39am

Thanks for this fantastic course; it has been a life saver to me!

In the file “lesson3-rossman.ipynb”, Just before the section titled “DL”, there is a line of code in it’s own cell that appears to be a bug:

val_idx=[0]

As a “beginner” I executed each cell, and this cell messed up my whole model from this point on. Apparently this is because it set my list of indexes of validation cases to 0 elements, which doesn’t seem to make sense. This is on the current distribution:

github.com

fastai/fastai/blob/master/courses/dl1/lesson3-rossman.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Structured and time series data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This notebook contains an implementation of the third place result in the Rossman Kaggle competition as detailed in Guo/Berkhahn's [Entity Embeddings of Categorical Variables](https://arxiv.org/abs/1604.06737).\n",
    "\n",
    "The motivation behind exploring this architecture is it's relevance to real-world application. Most data used for decision making day-to-day in industry is structured and/or time-series data. Here we explore the end-to-end process of using neural networks with practical structured data problems."
   ]
  },
  {
   "cell_type": "code",

This file has been truncated. show original

Could it be fixed if it is wrong?

Thank you,
Dana

asotov · July 16, 2018, 11:33am

I think, this line of code is a trap from Jeremy to catch students, who try to just hit Shift+Enter automatically.

MicPie · July 16, 2018, 4:36pm

The notebook is not meant to be executed from top to bottom as it includes several approaches.
Please search the forums as this is already discussed.

danaludwig · July 29, 2018, 8:18pm

You know, strangely some of my best learning is trouble-shooting things like this! Another question: where do people write rave reviews of this course? This is the best thing I’ve found on Deep Learning in about 18 months of looking! For the first time, I see a path out from being a perpetual newbie!

asotov · August 1, 2018, 7:26pm

I totally argee with you - Fast AI course open the door to the world of Deep Learning and Neural Networks for me. Before I start learning the course I think that NN is some kind of shamanism and tend to use classical machine learning methods. I were so excited, so I decided to share my insight with other guys who have been in my situation and prepare report for DUMP conference. DUMP is local Russian conference hosted in Ekaterinburg. Here is a video (in Russian) if somebody interested: https://www.youtube.com/watch?v=lJVAuFp5jc8 The name in english will be “How to ultimately get started with neural networks?”

Maybe, @rachel give us particular services where we can pay our respects.