Ex Machina is a one-day event for Machine Learning practitioners. We’re going to shed the hype and focus on the pragmatic stuff: how Machine Learning works in practice, how companies are using it, and how you can apply it today. We’ll have a few talks, and plenty of time for an open space discussion.
We'll start at 9:30 and finish at 18:00. We'll have talks in the morning, and plenty of time for an Open Space in the afternoon. We also have social events on the night before, if you're already in Bologna.
From deep learning to chatbots, Machine Learning is changing the world. Be a part of that change.*
* Well, a little bit of hype is OK.
Social Events on Thursday 24
Natural Language Processing has been used in machine learning for years. The explosion of social media has shifted the focus back to written natural language, and comes with many new ethical, social, and practical issues.
One such issue is the widespread diffusion of hate speech on social media platforms, and how to counteract it. In this talk, I will give an overview of the NLP approaches to hate speech detection, from the earliest methods based on dictionaries and support vector machines, to the new ones based on recurrent neural networks and word embeddings. I’ll show how to recognize hate speech from loosely labeled data. I will also showcase the first results of a current research project aimed at mapping the hate speech in Italy using data from Twitter.
We’ve had machine learning services in production at Quokky for about two years now. These services are the result of our collaboration with a large tax collection company, with the goal of automatic document classification using both textual & visual classifiers deployed on serverless cloud infrastructure.
In this talk, I'll share our experience moving towards ML solutions. We approached this challenge from the point of view of software/cloud engineers. We'll compare different ways of of classifying documents: the old-fashion C/Python way, classic cloud infrastructures, Tensorflow, hosted ML services, Serverless architectures.
Urban Wastewater Treatment Plants (WWTP) remove contaminants from municipal wastewater, that contains household sewage and industrial wastewater. They are dynamic and complex systems, because of the continuous variations of wastewater characteristics and the intrinsic complexity of biological processes.
To work effectively and efficiently, these processes require a proper level of dissolved oxygen to support aerobic bacteria. The level of oxygen is measured by DO sensors, that are exposed to high risk of failures because of the wastewater pollutants. In this talk, we tell how we used neural networks to identify sensor anomalies as soon as they happen.
Difficilmente la tua azienda è unica al mondo. Hai dei concorrenti, che magari forniscono gli stessi tuoi servizi. Ciononostante, molte aziende usano soluzioni AI custom invece di investigare le soluzioni che sono già disponibili.
L’Intelligenza Artificiale richiede grandi investimenti. Non innovare… riutilizza! Vedremo quali soluzioni cloud sono sul mercato, le tecnologie disponibili, e come possiamo riutilizzare ciò che già abbiamo. Descriveremo poi una nostra esperienza concreta che riguarda la news recommendation, mostrano come i servizi di AWS hanno semplificato il processo di deploy e la validazione della soluzione.
Cosa succede quando sei un editore e devi vagliare 9k commenti al giorno? Principalmente un aumento di caffeina nel sangue del tuo staff e una crescita esponenziale degli errori di moderazione. In questo talk illustreremo un case study su come abbiamo introdotto (e sperimentato) il machine learning (NLP) per la moderazione dei commenti a supporto dello staff editoriale.
We'll collect Open Space topics on the spot. If you wish, you can send an early topic proposal. A few ideas:
We’ll pick additional topics on the spot, as per the open space rules.