background preloader

The Unreasonable Effectiveness of Recurrent Neural Networks

The Unreasonable Effectiveness of Recurrent Neural Networks
There’s something magical about Recurrent Neural Networks (RNNs). I still remember when I trained my first recurrent network for Image Captioning. Within a few dozen minutes of training my first baby model (with rather arbitrarily-chosen hyperparameters) started to generate very nice looking descriptions of images that were on the edge of making sense. Sometimes the ratio of how simple your model is to the quality of the results you get out of it blows past your expectations, and this was one of those times. What made this result so shocking at the time was that the common wisdom was that RNNs were supposed to be difficult to train (with more experience I’ve in fact reached the opposite conclusion). We’ll train RNNs to generate text character by character and ponder the question “how is that even possible?” By the way, together with this post I am also releasing code on Github that allows you to train character-level language models based on multi-layer LSTMs. Recurrent Neural Networks Related:  AINeural Networks and AI

Recurrent Neural Networks Tutorial, Part 1 – Introduction to RNNs – WildML Recurrent Neural Networks (RNNs) are popular models that have shown great promise in many NLP tasks. But despite their recent popularity I’ve only found a limited number of resources that throughly explain how RNNs work, and how to implement them. That’s what this tutorial is about. As part of the tutorial we will implement a recurrent neural network based language model. I’m assuming that you are somewhat familiar with basic Neural Networks. What are RNNs? The idea behind RNNs is to make use of sequential information. A recurrent neural network and the unfolding in time of the computation involved in its forward computation. The above diagram shows a RNN being unrolled (or unfolded) into a full network. There are a few things to note here: You can think of the hidden state as the memory of the network. What can RNNs do? RNNs have shown great success in many NLP tasks. Language Modeling and Generating Text since we want the output at step to be the actual next word. Machine Translation .

CNTK, el nuevo paquete de herramientas de aprendizaje profundo de código abierto de Microsoft en GitHub - El blog de Windows para América Latina Microsoft ha comenzado a fabricar las herramientas que sus propios investigadores usan para acelerar los avances en inteligencia artificial que estén disponibles para un amplio grupo de desarrolladores, al lanzar su Paquete de Herramientas de Red Computacional en GitHub. Los investigadores desarrollaron este paquete de herramienta de código abierto, apodado CNTK, por necesidad. Xuedong Huang, jefe científico de habla en Microsoft, dijo que él y su equipo estaban ansiosos por realizar mejoras más rápidas en las formas en las que las computadoras entienden el habla, y cómo las herramientas con las que tenían que trabajar los retrasaban. Así que un grupo de voluntarios se prepararon para resolver este problema por sí solos, con ayuda de una solución casera que resaltó el rendimiento sobre todo lo demás. El esfuerzo rindió frutos. “El paquete de herramientas CNTK es mucho más eficiente que cualquier otra que hemos visto”, Huang dijo. Xuedong Huang (fotografía por Scott Eklund/Red Box Pictures

Understanding LSTM Networks -- colah's blog Posted on August 27, 2015 Recurrent Neural Networks Humans don’t start their thinking from scratch every second. As you read this essay, you understand each word based on your understanding of previous words. You don’t throw everything away and start thinking from scratch again. Traditional neural networks can’t do this, and it seems like a major shortcoming. Recurrent neural networks address this issue. Recurrent Neural Networks have loops. In the above diagram, a chunk of neural network, , looks at some input and outputs a value . These loops make recurrent neural networks seem kind of mysterious. An unrolled recurrent neural network. This chain-like nature reveals that recurrent neural networks are intimately related to sequences and lists. And they certainly are used! Essential to these successes is the use of “LSTMs,” a very special kind of recurrent neural network which works, for many tasks, much much better than the standard version. The Problem of Long-Term Dependencies Conclusion

Understanding Machine Learning Infographic Other Infographics Understanding Machine Learning Infographic Understanding Machine Learning Infographic We now live in an age where machines can teach themselves without human intervention. What It Is Machine learning (ML) deals with systems and algorithms that can learn from various data and make predictions. Theory The main goal of a learner is to generalize, and a learning machine able to do that can perform accurately on new or unforeseen tasks. History In the early days of AI, researchers were very interested in machines that could learn from data. How It Is Done Supervised ML – relies on data where the true label is indicated. Approaches There are over a dozen approaches employed in ML, Some of these include: Applications The importance of ML is that, since it’s data-driven, it can be trained to create valuable predictive models that can guide proper decisions and smart actions. Embed This Education Infographic on your Site or Blog!

Los intereses comerciales marcan el futuro de la inteligencia artificial | Ciencia El futuro de la inteligencia artificial genera muchos debates porque será decisiva en campos tan serios como la medicina, las guerras, el trabajo o incluso las relaciones humanas. Sin embargo, esos debates a menudo ignoran un asunto que sobrevuela a todos los demás: el desarrollo de las máquinas pensantes ha sido conquistado por empresas tecnológicas que están definiendo cómo será ese futuro. Compañías como Google, Facebook, Amazon, Microsoft, Apple e IBM fichan a los mejores expertos en inteligencia artificial de todo el mundo, esquilman departamentos universitarios enteros para cubrir sus necesidades, compran las empresas incipientes del sector y marcan el rumbo de la investigación con becas y ayudas. Así, un campo científico tan determinante como la inteligencia artificial puede estar volcado excesivamente en los intereses comerciales de estos negocios. ampliar foto No es solo compartir centro de trabajo con los mejores. Control sobre la academia

A Course in Machine Learning Google researchers teach AIs to see the important parts of images — and tell you about them | TechCrunch This week is the Computer Vision and Pattern Recognition conference in Las Vegas, and Google researchers have several accomplishments to present. They’ve taught computer vision systems to detect the most important person in a scene, pick out and track individual body parts and describe what they see in language that leaves nothing to the imagination. First, let’s consider the ability to find “events and key actors” in video — a collaboration between Google and Stanford. Footage of scenes like basketball games contain dozens or even hundreds of people, but only a few are worth paying attention to. The CV system described in this paper uses a recurrent neural network to create an “attention mask” for every frame, then track relevance of each object as time proceeds. Over time the system is able to pick out not only the most important actor, but potential important actors, and the events with which they are associated. Featured Image: Omelchenko/Shutterstock

The Difference Between AI, Machine Learning, and Deep Learning? This is the first of a multi-part series explaining the fundamentals of deep learning by long-time tech journalist Michael Copeland. Artificial intelligence is the future. Artificial intelligence is science fiction. Artificial intelligence is already part of our everyday lives. All those statements are true, it just depends on what flavor of AI you are referring to. For example, when Google DeepMind’s AlphaGo program defeated South Korean Master Lee Se-dol in the board game Go earlier this year, the terms AI, machine learning, and deep learning were used in the media to describe how DeepMind won. The easiest way to think of their relationship is to visualize them as concentric circles with AI — the idea that came first — the largest, then machine learning — which blossomed later, and finally deep learning — which is driving today’s AI explosion — fitting inside both. From Bust to Boom Over the past few years AI has exploded, and especially since 2015. Good, but not mind-bendingly great.

A 'Brief' History of Neural Nets and Deep Learning, Part 1 – Andrey Kurenkov's Web World This is the first part of ‘A Brief History of Neural Nets and Deep Learning’. Part 2 is here, and parts 3 and 4 are here and here. In this part, we shall cover the birth of neural nets with the Perceptron in 1958, the AI Winter of the 70s, and neural nets’ return to popularity with backpropagation in 1986. “Deep Learning waves have lapped at the shores of computational linguistics for several years now, but 2015 seems like the year when the full force of the tsunami hit the major Natural Language Processing (NLP) conferences.” -Dr. Christopher D. This may sound hyperbolic - to say the established methods of an entire field of research are quickly being superseded by a new discovery, as if hit by a research ‘tsunami’. I am in no capacity an expert on this topic. Let’s start with a brief primer on what Machine Learning is. Okay okay, enough definitions. Why have all this prologue with linear regression, since the topic here is ostensibly neural nets? of Donald Hebb. . .

AI just defeated a human fighter pilot in an air combat simulator Retired United States Air Force Colonel Gene Lee recently went up against ALPHA, an artificial intelligence developed by a University of Cincinnati doctoral graduate. The contest? A high-fidelity air combat simulator. And the Colonel lost. In fact, all the other AI’s that the Air Force Research Lab had in their possession also lost to ALPHA…and so did all of the other human experts who tried their skills against ALPHA’s superior algorithms. And did we mention ALPHA achieves superiority while running on a $US35 Raspberry Pi? Saying that Lee is experienced when it comes to aerial combat is a remarkable understatement. Yet, he was not successful in winning against ALPHA. "I was surprised at how aware and reactive it was. ALPHA makes decisions using a genetic fuzzy tree system, which is a subtype of fuzzy logic algorithms. The future of air combat UC grad and Psibernetix President and CEO Nick Ernest, David Carroll, and Gene Lee (seated).