GPT-3, a Giant Step for Deep Learning and NLP
Can intelligence emerge simply by training a big enough language model using lots of data? OpenAI tries to do so, using 175 billion parameters.
Another Datum
A blog by Yoel Zeldes
Latest posts
The Accessibility of GPT-2 - Text Generation and Fine-tuning
Text generation using GPT-2 is quite easy, using the right tools. Learn how to do it, as well as how to fine-tune the model on your own dataset.
Mixture of Variational Autoencoders - a Fusion Between MoE and VAE
An unsupervised approach to digit classification and generation.
TensorFlow — The Scope of Software Engineering
How to structure your TensorFlow graph like a software engineer.
Preparing for the Unexpected
How to apply your model to input it has never seen before.
Think your Data Different
Learn how node2vec works, and what kind of information it captures that word2vec doesn’t — includes case study.
How to Build Your Personal Brand as a Data Scientist
A couple of months ago I embarked on a journey to build my personal brand as a data scientist, and I want to share how I did it with you.
The Story of a Bad Train-Test Split
Splitting your dataset to train-test sets can sometimes be more complicated than one might expect.
TensorFlow Filesystem - Access Tensors Differently
The weird (but cool) way to access a tensorflow model by mounting it into a filesystem.
Variational Autoencoders Explained in Detail
Learn all the details needed to implement a variational autoencoder, code included.
How to Engineer Your Way Out of Slow Models
So you just finished designing that great neural network architecture. But how do you handle the fact it is slow?
Zooming Past the Competition
How to create an Augmented Reality app that allows a user to get content recommendations.
Variational Autoencoders Explained
Ever wondered how the Variational Autoencoder model works? Keep reading to find out.
Uncertainty for CTR Prediction: One Model to Clarify Them All
Learn how to handle uncertainty in recommender systems in a principled way using one unified model.
Recommender Systems: Exploring the Unknown Using Uncertainty
Learn what the exploration-exploitation tradeoff is, and how to use your model's uncertainty to explore new items in a wise manner.
Neural Networks from a Bayesian Perspective
Learn how to estimate model uncertainty in neural networks.
Using Uncertainty to Interpret your Model
Interpreting deep learning models is hard. Learn how to use uncertainty estimates to get insights about what the model has learned.
Neural Networks gone wild! They can sample from discrete distributions now!
Learn how to use Gumbel distribution to form a NN containing a discrete random component.
Deep Learning: Theory & Practice
Summary of TCE conference - "Deep Learning: Theory & Practice".
The Hitchhiker's Guide to Hyperparameter Tuning
Our implementation and usage of hyperparameter tuning at Taboola.
Word morphing
How to employ word2vec's embeddings and A* search algorithm to morph between words.
Gated Multimodal Units for Information Fusion
Learn how to train a neural network to use inputs from multiple different modalities using the GMU block.
Linear regression in the wild
Using linear regression when the dependant variables have measurement errors.
Location History Analysis
Analysis of Google Location History using Pandas time-series API.
Avoid committing junk
Have you ever accidentaly pushed temporary stuff to the remote git server? Then keep reading.
Pelican and GitHub Pages workflow
A simple workflow for writing blog posts using Pelican and GitHub Pages.
Detecting Anomalies in a SMART Way (Part Three)
Third part of a posts series about finding anomalous users.
Detecting Anomalies in a SMART Way (Part Two)
Second part of a posts series about finding anomalous users.
Detecting Anomalies in a SMART Way
Detecting anomalous user behavior is hard. Really hard. Here at Fortscale we show that with the right tools, it’s doable.
Who wants to be a coder
Learn how to use statistical tests to find out if a difference of means is significant, using the Coder Survey dataset as a case study.