BSC-CNS at NIPS2017, a top Machine Learning and Artificial Intelligence conference

2017-12-03T18:38:00+00:00 December 3rd, 2017|

BSC-CNS at NIPS2017, a top Machine Learning and Artificial Intelligence conference BSC-CNS will be present at the 31th Annual Conference on Neural Information Processing Systems (NIPS 2017), a machine learning and computational neuroscience conference that includes invited talks, demonstrations and oral and poster presentations. It covers topics ranging from deep learning and computer vision to cognitive science and reinforcement learning. NIPS is one of the top Machine Learning and Artificial Intelligence conferences in the world and has become the academic and industry AI conference. One paper will be presented at Machine Learning for Health workshop: "Detection-aided liver lesion segmentation using deep learning". In this paper we propose a method to segment the liver and its lesions from Computed Tomography (CT) scans using Convolutional Neural Networks (CNNs), that have proven good results in a variety of computer [...]

First steps with TensorFlow

2017-11-08T20:41:26+00:00 October 3rd, 2017|

First steps with TensorFlow In reality, this post was intended for my DLAI course's students, although I think it may be of interest to other students. I am going to share in this blog the teaching material that I am going to generate for the part of DLAI course that will cover the basic principles of Deep Learning from a computational perspective. TensorFlow TensorFlow is an open source software library for numerical computation using data flow graphs. Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. The flexible architecture allows you to deploy computation to one or more CPUs or GPUs in a desktop, server, or mobile device with a single API. Task 1:  Update DLAI [...]

Primeros pasos en Keras

2017-11-11T16:05:36+00:00 June 20th, 2017|

Keras es una librería de Python que proporciona de manera limpia y sencilla la creación de una gama de modelos de Deep Learning encima de otras librerías TensorFlow, Theano o CNTK. Keras fue desarrollado y es mantenido por François Chollet, un ingeniero de Google, y su código ha sido liberado bajo la licencia permisiva del MIT. Características básicas de Keras (*) Desconozco si fue la intención de François Chollet, pero personalmente valoro la austeridad y simplicidad que presenta este modelo de programación, sin adornos y maximizando la legibilidad. Permite expresar redes neuronales de una manera muy modular, considerando un modelo como una secuencia o un solo grafo. Una buena aproximación, a mi entender, porque los componentes de un modelo de Deep Learning son elementos discretos que [...]

First contact with Keras

2017-10-11T09:24:42+00:00 June 18th, 2017|

. . Keras is a Python library that provides a clean and convenient way to create a range of deep learning models on top of  powerful libraries such as TensorFlow, Theano (update about Theano) or CNTK. Keras was developed and maintained by François Chollet, a Google engineer and it is released under the permissive MIT license. Basic features of Keras (*) I value his austerity and simplicity, without frills approach and maximizing readability. It makes it possible to express neural networks in a very modular way, considering a model as a sequence or a graph alone. A good approximation for beginners, because the components of a Keras model are discrete elements that can be combined in arbitrary ways. New components are intentionally easy to add and modify within [...]

Caffe2: A new player on the scene of Deep Learning frameworks

2017-08-17T16:34:41+00:00 April 25th, 2017|

Few days ago Facebook announced Caffe2, a new open-source, cross-platform framework for deep learning. They say that Caffe2 is the successor to Caffe ( really?), the deep learning framework developed by Berkeley AI Research and community contributors. Caffe2’s GitHub page describes it as “an experimental refactoring of Caffe that allows a more flexible way to organize computation.” As my readers know, when appeared TensorFlow I decided to pay attention to it because it could change the scene of DL/AI frameworks. Now, we are in the same situation, Caffe2 could change the current scene that Francesc Sastre, one of my master students,  build for his master thesis:  "Frameworks popularity evolution in GitHub" No questions, right? Facebook launched Caffe2, an open-source deep learning framework made with expression, speed, and modularity in mind. It address the bottlenecks observed [...]

DeepMind moves to TensorFlow. Great!

2017-08-08T22:41:55+00:00 May 3rd, 2016|

This is great news for my book! It confirms that TensorFlow is a good choice for you too! For nearly four years, the open source Torch7 machine learning library has served as primary research platform at DeepMind.  Now it is time of TensorFlow. You can know more details in this post at Google Research Blog.  

New release of TensorFlow with distributed computing support

2017-08-08T22:45:33+00:00 April 13th, 2016|

Few hours ago Google announced his TensorFlow 0.8 that includes distributed computing support. As we already presented in this blog, distributed TensorFlow is powered by the high-performance gRPC library, which supports training on hundreds of machines in parallel according Google post. It complements the recent announcement of Google Cloud Machine Learning, which enables us to use the Google Cloud Platform. The post also announces that they have published a distributed trainer for the Inception image classification neural network in the TensorFlow models repository. The distributed trainer also enables us to scale out training using a cluster management system like Kubernetes from Google. Furthermore, once we have trained our model, we can deploy to production and speed up inference using TensorFlow Serving on Kubernetes. Beyond distributed Inception, the 0.8 release includes new libraries for defining our own distributed models. Using the distributed trainer, they trained the Inception network to 78% [...]

Calçotada season and the new Deep Learning book

2017-08-08T22:46:00+00:00 April 9th, 2016|

The calçotada season is coming to an end. Keep this excellent recipe from Tampa Bay Times for next year. Now it is time to read First contact with TensorFlow book. Already available a paper version, PDF version and Kindle version. Also this book is going to be freely available on-line in my web page (html version) next April 23th , Saint George's day  (Sant Jordi Day). This day is Barcelona's most romantic day of the year: St Jordi's is a day of Roses and Books. Come to see it! (*) I si no saps quin llibre regalar per St Jordi ...    Sant Jordi's day in all town and cities in Catalonia: #BooksAndRoses  

Aprender TensorFlow en Salamanca

2017-08-08T22:46:58+00:00 March 31st, 2016|

Hace unos días unos alumnos de la Facultad de Ciencias de la USAL pertenecientes al  capítulo de la ACM se pusieron en contacto conmigo a propósito del libro TensorFlow y en especial por una de sus ilustraciones.  Uno de sus intereses es el campo de la inteligencia artificial y las redes neuronales artificiales, así que desde que fundaron la asociación han venido organizando anualmente workshops con esta temática [1][2]. Ahora organizan un seminario introductorio al aprendizaje automático con redes neuronales y TensorFlow. Me han mostrado su github que han preparado para el curso y sin duda va a ser impresionante, tratando temas  como las LSTM que no contiene mi libro. Les pregunté que debería contar a mis alumnos para animarlos a asistir a este seminario en Salamanca. Aquí va: "La verdad es que estudiar en Salamanca es toda una [...]

¡Google ofrece más Machine Learning a los desarrolladores!

2017-08-08T22:47:09+00:00 March 24th, 2016|

Google acaba de anunciar en su blog nuevas herramientas de Machine Learning para  desarrolladores disponibles como un servicio más en su plataforma Google Cloud  a través de APIs. En el blog se indica que es exactamente la misma tecnología que está detrás de productos como Google Now o Google Photos, permitiendo a los desarrolladores construirse potentes modelos Machine Learning usando TensorFlow, además de ofrecer modelos preentrenados a través de Google Translate API,  Cloud Vision API o Google Cloud Speech API. Si juntamos esto con su servicio Cloud Dataproc que permite la gestión de procesos tanto de Hadoop como de Spark,  realmente los desarrollador e investigadores tenemos ahora mismo una potente y completa plataforma de procesado Big Data. Si tienen dos minutos les recomiendo este video insertado en su blog, para hacerse una idea a través de un simple robot realizado con una [...]

Distributed TensorFlow Has Arrived

2017-08-08T22:49:19+00:00 March 13th, 2016|

The landscape of Deep Learning was impacted in November, 2015, with the release of Google's TensorFlow, what is now the most popular open source machine learning library on Github by a wide margin. Some researchers showed their dissatisfaction with the project because the lack of distributed training capabilities (because such capabilities were directly alluded to in the accompanying whitepaper's title). However, the distributed TensorFlow has arrived, few week ago [*] Google announced an update to its deep learning library and TensorFlow now supports distributed training. The distributed version of TensorFlow is supported by gRPC, which is a high performance, open source RPC framework for inter-process communication (the same protocol used by TensorFlow Serving). Remember that the second most-starred machine learning project of Github is Scikit-learn, the de [...]

Google launched TensorFlow Serving

2017-08-09T12:13:22+00:00 February 18th, 2016|

Google launched TensorFlow Serving, that helps developers to take their TensorFlow machine learning models (and, even so, can be extended to serve other types of models) into production.  TensorFlow Serving is an open source serving system (written in C++) now available on GitHub under the Apache 2.0 license. What is the difference between TensorFlow and TensorFlow Serving?  While in TensorFlow is easier for the developers to build machine learning algorithms and train them for certain types of data inputs, TensorFlow Serving specializes in making these models usable in production environments.  The idea is that developers train their models using TensorFlow and then they use TensorFlow Serving’s APIs to react to input from a client. This allows developers to experiment with different models in a large scale that change over time based on real-world data, and maintain a [...]