Achieving faster execution with shorter compilation time can foster further diversity and innovation in neural networks. However, the current paradigm of executing neural networks either relies on hand-optimized libraries, traditional compilation heuristics, or very recently genetic algorithms and other stochastic methods. These methods suffer from frequent costly hardware measurements rendering them not only too time consuming but also suboptimal. As such, we devise a solution that can learn to quickly adapt to a previously unseen design space for code optimization, both accelerating the search and improving the output performance. This solution dubbed Chameleon leverages reinforcement learning whose solution takes fewer steps to converge, and develops an adaptive sampling algorithm that not only focuses on the costly samples (real hardware measurements) on representative points but also uses a domain-knowledge inspired logic to improve the samples itself. Experimentation with real hardware shows that Chameleon provides 4.45x speed up in optimization time over AutoTVM, while also improving inference time of the modern deep networks by 5.6%.

MRI images reconstructed from sub-sampled data using deep learning techniques often show a characteristic banding, which is particularly strong in low signal-to-noise regions of the reconstructed image. In this work, we propose the use of an adversarial loss that penalizes banding structures without requiring any human annotation. Our technique greatly reduces the appearance of banding, without requiring any additional computation or post-processing at reconstruction time. We report the results of a blind comparison against a strong baseline by a group of expert evaluators (board-certified radiologists), where our approach is ranked superior at banding removal with no statistically significant loss of detail.

Federated learning is a new machine learning framework which enables different parties to collaboratively train a model while protecting data privacy and security. Due to model complexity, network unreliability and connection in-stability, communication cost has became a major bottleneck for applying federated learning to real-world applications. Current existing strategies are either need to manual setting for hyper-parameters, or break up the original process into multiple steps, which make it hard to realize end-to-end implementation. In this paper, we propose a novel compression strategy called Residual Pooling Network (RPN). Our experiments show that RPN not only reduce data transmission effectively, but also achieve almost the same performance as compared to standard federated learning. Our new approach performs as an end-to-end procedure, which should be readily applied to all CNN-based model training scenarios for improvement of communication efficiency, and hence make it easy to deploy in real-world application without human intervention.

--The FloatSD technology has been shown to have excellent performance on low-complexity convolutional neural networks (CNNs) training and inference. In this paper, we applied FloatSD to recurrent neural networks (RNNs), specifically long short-term memory (LSTM). Moreover, the arithmetic precision for accumulations and the master copy of weights were reduced from 32 bits to 16 bits. We demonstrated that the proposed training scheme can successfully train several LSTM models from scratch, while fully preserving model accuracy. Finally, to verify the proposed method's advantage in implementation, we designed an LSTM neuron circuit and showed that it achieved significantly reduced die area and power consumption. I NTRODUCTION Recently, a great many studies have worked on proposing low-complexity, high-performance neural networks (NN) training and inference methods as well as their hardware architectures. The main objective is to reduce the huge computational and memory storage/access needed by popular modern NN models.

Human-machine interaction is increasingly dependent on speech communication. Machine Learning models are usually applied to interpret human speech commands. However, these models can be fooled by adversarial examples, which are inputs intentionally perturbed to produce a wrong prediction without being noticed. While much research has been focused on developing new techniques to generate adversarial perturbations, less attention has been given to aspects that determine whether and how the perturbations are noticed by humans. This question is relevant since high fooling rates of proposed adversarial perturbation strategies are only valuable if the perturbations are not detectable. In this paper we investigate to which extent the distortion metrics proposed in the literature for audio adversarial examples, and which are commonly applied to evaluate the effectiveness of methods for generating these attacks, are a reliable measure of the human perception of the perturbations. Using an analytical framework, and an experiment in which 18 subjects evaluate audio adversarial examples, we demonstrate that the metrics employed by convention are not a reliable measure of the perceptual similarity of adversarial examples in the audio domain.

--This paper presents a novel deep learning architecture for short term load forecasting of building energy loads. The architecture is based on a simple base learner and multiple boosting systems that are modelled as a single deep neural network. The architecture transforms the original multivariate time series into multiple cascading univariate time series. T ogether with sparse interactions, parameter sharing and equivariant representations, this approach makes it possible to combat against overfitting while still achieving good presentation power with a deep network architecture. The architecture is evaluated in several short-term load forecasting tasks with energy data from an office building in Finland. The proposed architecture outperforms state-of-the-art load forecasting model in all the tasks. Due to increasing utilization of renewables controlling the demand flexibility is becoming crucial part of the stabilization of smart grids. In this setting individual buildings are becoming key resources since buildings consume 32% of global final energy use [1].

Large-scale collections of electronic records constitute both an opportunity for the development of more accurate prediction models and a threat for privacy. To limit privacy exposure new privacy-enhancing techniques are emerging such as federated learning which enables large-scale data analysis while avoiding the centralization of records in a unique database that would represent a critical point of failure. Although promising regarding privacy protection, federated learning prevents using some data-cleaning algorithms thus inducing new biases. In this work we focus on the recurrent problem of duplicated records that, if not handled properly, may cause over-optimistic estimations of a model's performances. We introduce and discuss stratified cross-validation, a validation methodology that leverages stratification techniques to prevent data leakage in federated learning settings without relying on demanding deduplication algorithms.

We propose a flexible, co-creative framework bringing together multiple machine learning techniques to assist human users to efficiently produce effective creative designs. We demonstrate its potential with a perfume bottle design case study, including human evaluation and quantitative and qualitative analyses.

Usually two or more agents play a cooperative game where agents' goals are common but they access different information. In order to solve the given task agents have to share useful information with each other through a discrete bottleneck called the communication channel. The discrete symbols in the message do not have any a priori meaning but agents learn to cooperate by attributing meaning to the messages; a language protocol emerges as a byproduct of the training process. This emergent language serves only one goal: to complete the task successfully. One of the promises of this approach is to provide meaningful insights into the early stages of human language emergence as a result of cooperation.

Google Cloud Platform (GCP) today launched Deep Learning Containers, environments optimized for deploying and testing applications and services that utilize machine learning. Now in beta, GCP Deep Learning Containers works in the cloud and on-premises, making it possible to develop or prototype in both. Amazon introduced AWS Deep Learning Containers with Docker image support in March. Google plans for its Deep Learning Containers to "reach parity with all Deep Learning virtual machine types" in the future, according to a blog post sharing the news. The new service includes preconfigured Jupyter and Google Kubernetes Engine (GKE) clusters and launches with machine learning acceleration available from Nvidia GPUs, Intel CPUs, and other hardware.