The recent trend towards more programmable switching hardware in data centers opens up new possibilities for distributed applications to leverage in-network computing (INC). Literature so far has largely focused on individual application scenarios of INC, leaving aside the problem of coordinating usage of potentially scarce and heterogeneous switch resources among multiple INC scenarios, applications, and users. Alas, the traditional model of resource pools of isolated compute containers does not fit an INC-enabled data center. This paper describes HIRE, a holistic INC-aware resource manager which allows for server-local and INC resources to be coordinated in unison. HIRE introduces a novel flexible resource (meta-)model to address heterogeneity and resource interchangeability, and includes two approaches for INC scheduling: (a) retrofitting existing schedulers; (b) designing a new one. For (a), HIRE presents a retrofitting API and demonstrates it with four state-of-the-art schedulers. For (b), HIRE proposes a flow-based scheduler, cast as a min-cost max-flow problem, where a unified cost model is used to integrate the different costs. Experiments with a workload trace of a 4000 machine cluster show that HIRE makes better use of INC resources by serving 8–30% more INC requests, while simultaneously reducing network detours by 20% and reducing tail placement latency by 50%.

The recent trend towards more programmable switching hardware in data centers opens up new possibilities for distributed applications to leverage in-network computing (INC). Literature so far has largely focused on individual application scenarios of INC, leaving aside the problem of coordinating usage of potentially scarce and heterogeneous switch resources among multiple INC scenarios, applications, and users. The traditional model of resource pools of isolated compute containers does not fit an INC-enabled data center.

LiDAR sensors are crucial in automotive perception for accurate object detection. However, LiDAR data is hard to interpret for humans and consequently time-consuming to label. Whereas camera data is easy interpretable and thus, comparably simpler to label. Within this work we present a transductive transfer learning approach to transfer the knowledge for the object detection task from images to point cloud data. We propose a multi-modal adversarial Auto Encoder architecture which disentangles uni-modal features into two groups: common (transferable) features, and complementary (modality-specific) features. This disentanglement is based on the hypothesis that a set of common features exist. An important point of our framework is that the disentanglement is learned in an unsupervised manner. Furthermore, the results show that only a small amount of multi-modal data is needed to learn the disentanglement, and thus to transfer the knowledge between modalities. As a result we our experiments show that training with 75% less data of the KITTI objects, the classification accuracy achieved is of 71.75%, only 3.12% less than when using the full data set. The implications of these findings can have great impact in perception pipelines based on LIDAR data.

For autonomous driving applications it is critical to know which type of road users and road side infrastructure are present to plan driving manoeuvres accordingly. Therefore autonomous cars are equipped with different sensor modalities to robustly perceive its environment. However, for classification modules based on machine learning techniques it is challenging to overcome unseen sensor noise. This work presents an object classification module operating on unsupervised learned multi-modal features with the ability to overcome gradual or total sensor failure. A two stage approach composed of an unsupervised feature training and a uni-modal and multimodal classifiers training is presented. We propose a simple but effective decision module switching between uni-modal and multi-modal classifiers based on the closeness in the feature space to the training data. Evaluations on the ModelNet 40 data set show that the proposed approach has a 14% accuracy gain compared to a late fusion approach operating on a noisy point cloud data and a 6% accuracy gain when operating on noisy image data.