Rafael Ballestas

Here's a reflection on the need to represent code before actually feeding it into neural network based encoders, such as code2vec, word2vec, and code2seq.

Here we talk about Code2seq, which differs in adapting neural machine translation techniques to the task of mapping a snippet of code to a sequence of words.

Here is a tutorial on the usage of code2vec to predict method names, determine the accuracy of the model, and exporting the corresponding vector embeddings.

Here we discuss code2vec relation with word2vec and autoencoders to grasp better how feasible it is to represent code as vectors, which is our main interest.

This post is an overview of word2vec, a method for obtaining vectors that represent natural language in a way that is suitable for machine learning algorithms.

This post is a high-level review of our previous discussion concerning machine learning techniques applied to vulnerability discovery and exploitation.

Here is a simple attempt to define a vulnerability classifier using categorical encoding and a basic neural network with a single hidden layer.

In this post, we begin to tackle why vectors are the most appropriate representation for data as input to machine learning algorithms.