JAX Archives - AI Chronicles: When Machines Do It Better

word2vec cbow negative sampling feature image

Journey Through Embeddings: CBOW with negative sampling

In previous post, we explored the fundamentals of Word2Vec and implemented the Continuous Bag of Words (CBOW) model, which efficiently predicts target words based on their surrounding context. While CBOW is powerful, scaling it to large vocabularies can be computationally challenging due to the softmax function. This is where negative sampling comes in—a clever optimization […]

Journey Through Embeddings: CBOW with negative sampling Read More »

Journey Through Embeddings: Word2vec CBOW

Word embeddings have revolutionized natural language processing by transforming text into meaningful numerical representations that capture the relationships between words. In this blog, we focus on Word2Vec, a foundational technique in embedding methods, with a particular emphasis on the Continuous Bag of Words (CBOW) model. CBOW learns embeddings by predicting a target word from its

Journey Through Embeddings: Word2vec CBOW Read More »

Unlocking CNNs with JAX: Comprehensive guide

Let’s resume our hunt into the “Unlocking in JAX” series and let’s start with the type of neural networks which changed the way of how to deal with vision in general and application in real-world cases. Today, we dive into Convolutional Neural Networks (CNNs), exploring their unique ability to capture spatial hierarchies in images. By

Unlocking CNNs with JAX: Comprehensive guide Read More »

Unlocking Neural Networks with JAX: optimizers with optax

In our previous exploration within the “Unlocking in JAX” series, we covered the foundational concepts of neural networks and their implementation using JAX. Building on that knowledge, this post, shifts focus to a crucial component that significantly enhances learning efficiency: optimizers. Optax, a gradient processing and optimization library tailored for JAX, provides the tools necessary

Unlocking Neural Networks with JAX: optimizers with optax Read More »

Unlocking Neural Networks with JAX

Welcome back to our Unlocking with JAX series! Today, we’re getting hands-on with neural networks by building a multilayer perceptron (MLP) using JAX. JAX sits perfectly between the lower-level, details of CUDA and the higher-level abstractions offered by frameworks like Keras, offering both clarity and control. This balance helps immensely in understanding the inner workings

Unlocking Neural Networks with JAX Read More »