In the era of the Internet of Things (IoT), conventional cloud-based solutions struggle to handle the huge amount, high velocity, and heterogeneity of data generated at the network edge. In this context, the edge-to-cloud compute continuum has …
This post shows how to build an unsupervised deep learning model for digit generation by leveraging a convolutional variational autoencoder trained on the MNIST dataset of handwritten digits using Keras+Tensorflow.
In this post I show how to leverage BERT, a transformer-based language representation model, in order to identify the personality type of users based on their writing style and the content of their posts, according to the Myers-Briggs indicator (MBTI).
In what follows, I'll show how to fine-tune a BERT classifier, using Huggingface and Keras+Tensorflow, for dealing with two different text classification problems. The first consists in detecting the sentiment (*negative* or *positive*) of a movie review, while the second is related to the classification of a comment based on different types of toxicity, such as *toxic*, *severe toxic*, *obscene*, *threat*, *insult* and *identity hate*.
This post is dedicated to the development of an artificial intelligence application capable of identifying the emotions expressed through the voice in spoken language. The classification model focuses on seven different emotions (*anger*, *boredom*, *disgust*, *fear*, *happiness*, *sadness*, *neutral*) and is enhanced with the attention mechanism.
In what follows, I'll show how to build a dog breed classifier based on Convolutional Neural Networks, which focuses on two particular breeds: Chihuahua and Pug. In order to cope with the small amount of traning data, the model exploits three main techniques: real time data augmentation, Transfer Learning and fine tuning.