BLOG POSTS
NBA Video Dataset
Introducing my new NBA play-by-play video dataset...
In this blog post, I will be discussing the contents, download procedure, and intended use for my new NBA play-by-play dataset. In particular, this post will touch on the format of the downloaded dataset, types of labels for each event clip, and step-by-step instructions for acquiring and ultimately using the dataset on AWS.
Triplet Loss
Discussion on the under-appreciated triplet loss...
In typical deep learning and deep RL papers, you will commonly find references to cross-entropy and mean-squared error loss functions. However, though seldom used, the triplet loss function has show inordinate promise as an alternative in domains such as facial recognition and generating action embeddings. This article will define the triplet loss function, look at some of its use cases, and then present custom tools for training a model using the triplet loss function in Keras.
Distributed Training
Setting-up and using HPC cluster for DL on AWS...
As both DL models and datasets are continually getting larger, distributed training is becoming increasingly important for reducing training times to a manageable range (e.g. hours vs days and days vs weeks). For that reason, this tutorial is designed to help people interested in training complex models on larger datasets quickly get started on using a cluster of AWS instances.