Learning Curves
Learning Curves
1h 2m

Move on from what you learned from studying the principles of recurrent neural networks, and how they can solve problems involving sequencing, with this cohesive course on Improving Performace. Learn to improve the performance of your neural networks by starting with learning curves that allow you to answer the right questions. This could be needing more data or, even, building a better model to improve your performance.

Further into the course, you will explore the fundamentals around bash normalization, drop-out, and regularization.

This course also touches on data augmentation and its ability to allow you to build new data from your starting training data, culminating in hyper-parameter optimization. This is a tool to that aids in helping you to decide how to tune the external parameters of your network.

This course is made up of 13 lectures and three accompanying exercises. This Cloud Academy course is in collaboration with Catalit.

 Learning Objectives

  • Learn how to improve the performance of your neural networks.
  • Learn the skills necessary to make executive decisions when working with neural networks.

Intended Audience


Hello and welcome to this video about learning curves. In this video, we will talk about learning curves. A trained machine learning model will have a certain performance over a test set. If we are not satisfied with its value, we may want to improve the performance. At which point, it's legitimate to ask the question, should we collect more data or should we try a different model? The learning curve answers this question. Suppose we want to train an algorithm to distinguish cats from dogs and suppose we have 1,500 data points. These are labeled pictures of cats and dogs. We trained the model and we get 95% accuracy on a test set of 300 pictures. Now the question is, should we collect more pictures or think of a better model? The way to know the answer to the question is to do the following. 

First, we isolate the 300 test pictures from the rest of the data and then we train the model using only a fraction of the remaining data, for example 100 pictures. We calculate the score on the training set and we calculate the score on the test set. Since the training set is small, we expect the model to perform really well on it, possibly over fitting, while the model will perform quite poorly on the test set. Then we gradually take more pictures from the training set and train the model on a larger and larger fraction of the training set. With more training examples, the model should learn to generalize better and the test score should increase. We proceed like this until we've used all our training data. At this point, two cases are possible. If it looks like the test performance stopped increasing with the size of the training set, we probably reached the maximum performance of our model. 

In this case, we should invest time in looking for a better model to improve performance. In the other case, it could seem that the test area will continue to decrease if only we had access to more training data. If that's the case, we should probably go out looking for more labeled data first and then worry about changing the model. So now you know how to answer the big question, more data or better model? Use a learning curve. In this video, we've introduced the tool of the learning curve which is a very useful tool to answer the question whether you need more data or a better model to improve the performance of your model. Thank you for watching and see you in the next video.

About the Author
Learning Paths

I am a Data Science consultant and trainer. With Catalit I help companies acquire skills and knowledge in data science and harness machine learning and deep learning to reach their goals. With Data Weekends I train people in machine learning, deep learning and big data analytics. I served as lead instructor in Data Science at General Assembly and The Data Incubator and I was Chief Data Officer and co-­founder at Spire, a Y-Combinator-­backed startup that invented the first consumer wearable device capable of continuously tracking respiration and activity. I earned a joint PhD in biophysics at University of Padua and Université de Paris VI and graduated from Singularity University summer program of 2011.