GABC2018: Invert desktop entropy with #AI by Marco Farina

How many times have you ever been unable to remember exactly the name of a file or its content but being sure you have saved it in that folder? Still today, despite the evolution of the internal computer search systems, the subdivision of files into folders significantly simplifies the search for information by leveraging human visual memory, which is often the most developed. The only downside is the struggle of keeping each file correctly archived and therefore having to retrace each time the entire path that leads to the appropriate folder. But today, thanks to the combined use of Microsoft Cognitive Services and Azure Machine Learning Studio, it is possible to develop an intelligent sorter to be left on your desktop that automatically places in the right folder any file dragged on it. To be successful, it employs Microsoft technologies to map the content of each file and with them train a classification algorithm that quickly becomes reliable to classify files better than people. This of course works very well with existing folders, but when do I need to create a new folder? The answer within speech.

Transcript

1. Invert desktop entropy with #AI Global Azure Bootcamp – Zurich 21 April 2018
2. Logol – The Logical Path to Your Goal A new way through the future MBA, serving as a strategic technology leader, Marco Farina is a professional with performance-based leadership and knowledge of private and public sector business and IT systems. His experience includes leading information services for large-scale organizations while serving different organizations. His experience includes directing systems and programs for financial market and e-procurement, telecommunications and healthcare. Marco Farina – Founder & CEO Mathematician, after a career in strategic, management and financial consulting, Alessandro Maserati was in charged of the development of innovative IT systems for important Swiss companies. Within Logol's Artificial Intelligence team he is in charge of defining strategic objectives and designing neural networks. Alessandro Maserati – #AI Manager Engineer with a brilliant university career, after being a founding member of a Social Commerce startup, Nicola Casamassima has enriched his professional experience taking part to the development of innovative FinTech systems for a Swiss leader of a field of finance. Now he is a Logolian AI expert that leads the development of strategic solutions using neural networks. Nicola Casamassima – #AI Associate Main Address Corso Bello 2, Mendrisio (Ticino) Phone Number 0041 (0) 91 210 5827 Company E-mail Address info@logol.com Fax Number Are you kidding? 2
3. A common problem www.logol.com
4. The Solution: Learn The User’s Logics Reorganize new files using machine learning www.logol.com 4 Movie Root Sport Relax Movie 1 Movie 2 Rugby Books Painting Golf Horror Water Music Movie 3 Rock Swim Azure ML Movie Root Sport Relax Horror Water Music Movie 1 Movie 2 Rugby Books Painting Golf Movie 3 Rock Swim
5. The AI Context Machine learning – a subset of AI www.logol.com 5
6. The Machine Learning World www.logol.com 6
7. A Process Based on DL and ML Techniques A machine learning core encapsulated in a smart framework www.logol.com 7 Step 1 Step 2 Step 3 Step 4 Step 5 Step 6 Extract n-grams from text Identify new file fingerprints Suggest new file position Extract text from files Identify folders fingerprints Identify new file distance from folders
8. The Machine Learning Supervised Structure www.logol.com 8
9. Choose the right model Two Class Bayes Point Machine or Multiclass Neural Network www.logol.com 9 Multiclass Neural Network Use a neural network model that can be used to predict a target that has multiple values. Two-Class Bayes Point Machine Efficiently approximates the theoretically optimal Bayesian average of linear classifiers (in terms of generalization performance) by choosing one "average" classifier, the Bayes Point.
10. Let’s go see a demo On Azure Machine Learning Studio
11. Thanks for Your Attention www.logol.com