Pedro Larrañaga is Full Professor in Computer Science and Artificial Intelligence at the Universidad Politécnica de Madrid (UPM) since 2007, where he co-leads the Computational Intelligence Group. He received the MSc degree in mathematics (statistics) from the University of Valladolid and the PhD degree in computer science from the University of the Basque Country (excellence award). Before moving to UPM, his academic career was developed at the University of the Basque Country (UPV-EHU) at several faculty ranks: Assistant Professor (1985-1998), Associate Professor (1998-2004) and Full Professor (2004-2007). He earned the habilitation qualification for Full Professor in 2003. Professor Larrañaga has served as Expert Manager of Computer Technology area at the Deputy Directorate of research projects of the Spanish Ministry of Science and Innovation (2007-2010).

We propose a simple yet effective technique to simplify the training and the resulting model of neural networks. In back propagation, only a small subset of the full gradient is computed to update the model parameters. The gradient vectors are sparsified in such a way that only the top-k elements (in terms of magnitude) are kept. As a result, only k rows or columns (depending on the layout) of the weight matrix are modified, leading to a linear reduction in the computational cost. Based on the sparsified gradients, we further simplify the model by eliminating the rows or columns that are seldom updated, which will reduce the computational cost both in the training and decoding, and potentially accelerate decoding in real-world applications. Surprisingly, experimental results demonstrate that most of time we only need to update fewer than 5% of the weights at each back propagation pass. More interestingly, the accuracy of the resulting models is actually improved rather than degraded, and a detailed analysis is given. The model simplification results show that we could adaptively simplify the model which could often be reduced by around 9x, without any loss on accuracy or even with improved accuracy. The codes, including the extension, are available at https://github.com/lancopku/meSimp

We introduce Continual Learning via Neural Pruning (CLNP), a new method aimed at lifelong learning in fixed capacity models based on neuronal model sparsification. In this method, subsequent tasks are trained using the inactive neurons and filters of the sparsified network and cause zero deterioration to the performance of previous tasks. In order to deal with the possible compromise between model sparsity and performance, we formalize and incorporate the concept of graceful forgetting: the idea that it is preferable to suffer a small amount of forgetting in a controlled manner if it helps regain network capacity and prevents uncontrolled loss of performance during the training of future tasks. CLNP also provides simple continual learning diagnostic tools in terms of the number of free neurons left for the training of future tasks as well as the number of neurons that are being reused. In particular, we see in experiments that CLNP verifies and automatically takes advantage of the fact that the features of earlier layers are more transferable. We show empirically that CLNP leads to significantly improved results over current weight elasticity based methods.

Robots need to understand their environment to perform their task. If it is possible to pre-program a visual scene analysis process in closed environments, robots operating in an open environment would benefit from the ability to learn it through their interaction with their environment. This ability furthermore opens the way to the acquisition of affordances maps in which the action capabilities of the robot structure its visual scene understanding. We propose an approach to build such affordances maps by relying on an interactive perception approach and an online classification. In the proposed formalization of affordances, actions and effects are related to visual features, not objects, and they can be combined. We have tested the approach on three action primitives and on a real PR2 robot.

In this paper, we aim at providing an introduction to the gradient descent based optimization algorithms for learning deep neural network models. Deep learning models involving multiple nonlinear projection layers are very challenging to train. Nowadays, most of the deep learning model training still relies on the back propagation algorithm actually. In back propagation, the model variables will be updated iteratively until convergence with gradient descent based optimization algorithms. Besides the conventional vanilla gradient descent algorithm, many gradient descent variants have also been proposed in recent years to improve the learning performance, including Momentum, Adagrad, Adam, Gadam, etc., which will all be introduced in this paper respectively.

Multiagent Reinforcement Learning (RL) solves complex tasks that require coordination with other agents through autonomous exploration of the environment. However, learning a complex task from scratch is impractical due to the huge sample complexity of RL algorithms. For this reason, reusing knowledge that can come from previous experience or other agents is indispensable to scale up multiagent RL algorithms. This survey provides a unifying view of the literature on knowledge reuse in multiagent RL. We define a taxonomy of solutions for the general knowledge reuse problem, providing a comprehensive discussion of recent progress on knowledge reuse in Multiagent Systems (MAS) and of techniques for knowledge reuse across agents (that may be actuating in a shared environment or not). We aim at encouraging the community to work towards reusing all the knowledge sources available in a MAS. For that, we provide an in-depth discussion of current lines of research and open questions.

For those of us who worry that Facebook may have serious boundary issues when it comes to the personal information of its users, Mark Zuckerberg's recent comments at Harvard should get the heart racing. Zuckerberg dropped by the university last month ostensibly as part of a year of conversations with experts about the role of technology in society, "the opportunities, the challenges, the hopes, and the anxieties." His nearly two-hour interview with Harvard law school professor Jonathan Zittrain in front of Facebook cameras and a classroom of students centered on the company's unprecedented position as a town square for perhaps 2 billion people. To hear the young CEO tell it, Facebook was taking shots from all sides--either it was indifferent to the ethnic hatred festering on its platforms or it was a heavy-handed censor deciding whether an idea was allowed to be expressed. Zuckerberg confessed that he hadn't sought out such an awesome responsibility.

We are happy to announce that Microsoft and Intel are partnering to bring optimized deep learning frameworks to Azure. These optimizations are available in a new offering on the Azure marketplace called the Intel Optimized Data Science VM for Linux (Ubuntu). Over the last few years, deep learning has become the state of the art for several machine learning and cognitive applications. Deep learning is a machine learning technique that leverages neural networks with multiple layers of non-linear transformations, so that the system can learn from data and build accurate models for a wide range of machine learning problems. Computer vision, language understanding, and speech recognition are all examples of deep learning at play today.

I did this setup on Raspberry Pi3 and Pi3 . Probably, for Pi zero this setup needs an extra step. This process takes so much time (my case almost an hour). This process also takes so much time (my case almost an hour). Eventually, let's try run image classifier on Pi This could take so much time.

Use Python & Keras to do 24 Projects - Recognition of Emotions, Age, Gender, Object Detection, Segmentation, Face Aging Master Computer Vision using Deep Learning in Python. You'll be learning to use the following Deep Learning frameworks. In this course, you will discover the power of Computer Vision in Python, and obtain skills to dramatically increase your career prospects as a Computer Vision developer.Computer vision applications involving Deep Learning are booming! Having Machines that can'see' will change our world and revolutionize almost every industry out there. Machines or robots that can see will be able to: Perform surgery and accurately analyze and diagnose you from medical scans.