Don t miss out on any of the newest findings in these rapidly growing fields of research presented at the Deep Learning (DLSS) and Reinforcement Learning (RLSS) Summer School.

Since machine learning is a panacea, your company should be able to use it profitably, right? Let's discuss whether you have what it takes to harness artificial intelligence -- and how you could get to that point if you're not yet there. To begin with, do you know what you want to predict or detect? Do you have enough data to analyze to build predictive models? Do you have the people and tools you need to define and train models?

In recent years we've seen a pretty incredible explosion in CV (computer vision) enabled devices, and the software that powers them. It's not uncommon to see autonomous robots with multiple high definition cameras using sophisticated CV algorithms for guidance, even on low-cost hobbyist robots. But, that capability comes at a pretty big processing expense. Those HD videos contain a lot of pixels, and even with techniques for optimization that make processing the video more efficient, it still takes a lot processing power to do all of the complex analysis we want our robots to do. To try try and jump ahead of these increasing demands, Intel has just announced Movidius Myriad X, which is the first vision processing unit (VPU) that has a dedicated Neural Compute Engine built into the chip.

It's been almost a year since Intel scooped up AI and computer vision chip-maker Movidius. By the time of its takeover, the company had already crammed its Myriad 2 processor into drones, cameras, and USB sticks -- making it a good fit for Intel's beyond-the-PC strategy. The newly-unveiled successor to that chip will continue in the same vein. Only, this next-gen beast is the first to pack a "Neural Compute Engine." What that essentially means is that it has some pretty powerful deep learning capabilities.

I recently came across an ICML'17 paper "Attentive Recurrent Comparators" which proposes a simple yet powerful model for data efficient learning. The paper presents the first super-human One-shot Classification performance on the Omniglot dataset using only raw pixel information! In this blog post I am going to present my understanding of the main ideas of the paper. The authors of the paper also released an implementation using Theano. But I found the implementation a bit difficult to follow.

If the barrier to precision medicine is data handling, then artificial intelligence (AI) may be the logical solution.

Let start at the beginning; back in the 50's we dreamt about robots making our tea (AI) and what we could do with our spare time, fuelled by the space race and science fiction it was a heady dream. Until that is the 80's, where we started to'teach' machines simple tasks (Machine Learning), the more times a machine completed a task the more it learnt and improved. We've now reached Deep Learning, where through the improvement of data storage, processing power and network speed algorithms can be used so that software can train itself through multiple iterations and vast amounts of data.

Towards making systems forget with machine unlearning.

I was wrong to say that Intel (INTC) doesn't need GPUs to compete with Nvidia (NVDA) on artificial intelligence/deep learning computing. Further research told me that along with FPGA (Field Programmable Field Gate Array), there's an embedded Intel Processor Graphics for deep learning inference. It's a new concept that was discussed by Intel only last May. Nvidia's GPU can be the Training Engine for deep learning computers. Intel's FPGAs and embedded Processor Graphics could be the go-to hardware accelerators for inference computing.

Pytorch-C is a simple C 11 library which provides a Pytorch-like interface for building neural networks and inference (so far only forward pass is supported). The library respects the semantics of torch.nn Models from pytorch/vision are supported and can be easily converted. The library heavily relies on an amazing ATen library and was inspired by cunnproduction. The structure of the project and CMake will be changed in a future, as it is not optimal now.