Teaching Machines to Learn http://emanueleghelfi.github.io Object Detection: A Review In this article, I review some 2D Object Detection papers. R-CNN R-CNN (Girshick et al., 2014) (Regions with CNN features) presents a first, preliminary, approach to Object Detection. Figure 1: R-CNN Pipeline: Extract around 2000 bottom up region proposals from the input image Warp regions to... Sat, 29 Jan 2022 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2022/01/29/object_detection.html http://emanueleghelfi.github.io/blog/2022/01/29/object_detection.html I’m really proud to announce that “The Reinforcement Learning Workshop” is finally published! Packt Interactive Amazon Starting with an introduction to RL, you’ll be guided through different RL environments and frameworks. You’ll learn how to implement your own custom environments and use OpenAI baselines to run RL algorithms. Once you’ve... Fri, 03 Jul 2020 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2020/07/03/rl-workshop.html http://emanueleghelfi.github.io/blog/2020/07/03/rl-workshop.html I attented the course Regularization Methods for Machine Learning (RegML 2020). Topics: Regularization Methods: Penalization, Projection, Early Stopping. Kernel Methods Regularization in Multi-Task Learning Sparsity Geometric Deep Learning Representation Learning Here you can find my notes about the lessons. There can be errors, imprecisions and so on. They are still... Fri, 03 Jul 2020 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2020/07/03/regml.html http://emanueleghelfi.github.io/blog/2020/07/03/regml.html I went through the last Facebook paper (Carion et al., 2020), which proposes a new architecture, DETR, for object detection and segmentation based on Transformers. To better understand the paper, I decided to review the basic principles of Transformers and Attention. I am not pretending to write the best article... Fri, 06 Mar 2020 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2020/03/06/attention.html http://emanueleghelfi.github.io/blog/2020/03/06/attention.html .PDF Web Version Here I share a brief summary of the theory behind Machine Learning. These notes were taken during my M.Sc. (2016/2017) following the Machine Learning lectures. They might contain errors, they can be imprecise, they may contain typos. You can contact me if you spot errors, if you... Tue, 03 Sep 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/09/03/machine-learning-notes.html http://emanueleghelfi.github.io/blog/2019/09/03/machine-learning-notes.html Great news! The Zuru ML Team (me, Federico Di Mattia, Paolo Galeone, Michele De Simoni) will present the tutorial: “Deep Diving into GANs: From Theory to Production with TensorFlow 2.0” at EuroSciPy in Bilbao, the 2nd of September 2019. This tutorial builds upon the PyConX (see also here for more... Fri, 02 Aug 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/08/02/euroscipy-gans.html http://emanueleghelfi.github.io/blog/2019/08/02/euroscipy-gans.html AshPy is a TensorFlow 2.0 library for (distributed) training, evaluation, model selection, and fast prototyping. It is designed to ease the burden of setting up all the nuances of the architectures built to train complex custom deep learning models. AshPy, why? Ashpy provides already prepared trainers for classifier models and... Thu, 01 Aug 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/08/01/ashpy.html http://emanueleghelfi.github.io/blog/2019/08/01/ashpy.html The paper Toward Multimodal Image-to-Image Translation (Zhu et al., 2017) faces the problem of multimodal image-to-image translation in a principled way. The same problem is faced in (Isola et al., 2016) (pix2pix) but in a not-so-effective manner (enabling Dropout at inference time). In this paper, the authors encourage a bijection... Wed, 26 Jun 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/06/26/multimodal-image-to-image-translation.html http://emanueleghelfi.github.io/blog/2019/06/26/multimodal-image-to-image-translation.html This format is deliberately inspired by Adrian Colyer, Denny Britz and Daniel Takeshi. The series Paper Notes will contain the notes about papers I’ve read. This can be useful for people trying to understand better some mysterious paper and it can be very useful for me to obtain a deeper... Mon, 24 Jun 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/06/24/paper-notes.html http://emanueleghelfi.github.io/blog/2019/06/24/paper-notes.html Paper Poster Slides Code Most of the problems tackled by Reinforcement Learning are typically modeled as Markov Decision Processes in which the environment is considered a fixed entity and cannot be controlled. Nevertheless, there exist several real-world examples in which a partial control on the environment can be exercised by... Wed, 22 May 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/05/22/rl-conf-mdp.html http://emanueleghelfi.github.io/blog/2019/05/22/rl-conf-mdp.html Me and my colleagues (Federico Di Mattia, Paolo Galeone, Michele De Simoni) held the training: “Deep Diving into GANs: From Theory To Production” at PyCon X in Florence. GANs are the new hottest topic in the ML arena; however, they present a challenge for the researchers and the engineers alike.... Mon, 06 May 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/05/06/pyconx-gans.html http://emanueleghelfi.github.io/blog/2019/05/06/pyconx-gans.html Paper Slides In this article we present our approach for the NIPS 2017 ”Learning To Run” challenge. The goal of the challenge is to develop a controller able to run in a complex environment, by training a model with Deep Reinforcement Learning methods. We follow the approach of the team... Thu, 28 Mar 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/03/28/learning-to-run.html http://emanueleghelfi.github.io/blog/2019/03/28/learning-to-run.html Managing Machine Learning experiments is usually painful. The usual workflow when approaching a problem using Machine Learning tools is the following. You study the problem, define a possible solution, then you implement that solution and measure its quality. Often the solution depends on some parameters, and we refer to the... Tue, 12 Feb 2019 08:00:00 +0000 http://emanueleghelfi.github.io/blog/2019/02/12/sacred.html http://emanueleghelfi.github.io/blog/2019/02/12/sacred.html