Recent research on predicting the binding affinity between drug molecules and proteins use representations learned, through unsupervised learning techniques, from large databases of molecule SMILES and protein sequences. While these representations have significantly enhanced the predictions, they are usually based on a limited set of modalities, and they do not exploit available knowledge about existing relations among molecules and proteins. Our study reveals that enhanced representations, derived from multimodal knowledge graphs describing relations among molecules and proteins, lead to state-of-the-art results in well-established benchmarks (first place in the leaderboard for Therapeutics Data Commons benchmark ``Drug-Target Interaction Domain Generalization Benchmark", with an improvement of 8 points with respect to previous best result). Moreover, our results significantly surpass those achieved in standard benchmarks by using conventional pre-trained representations that rely only on sequence or SMILES data. We release our multimodal knowledge graphs, integrating data from seven public data sources, and which contain over 30 million triples. Pretrained models from our proposed graphs and benchmark task source code are also released.

We study the problem of operating a quantum switch with memory constraints. In particular, the switch has to allocate quantum memories to clients to generate link-level entanglements (LLEs), and then use these to serve end-to-end entanglements requests. The paper’s main contributions are (i) to characterize the switch’s capacity region and study how it scales with respect to the number of quantum memories and probability of successful LLEs and (ii) to propose a memory allocation policy that is throughput optimal. In addition, when the requests are bipartite and the LLE attempts are always successful, we show that the proposed policy has polynomial time complexity. We evaluate the proposed policy numerically and illustrate its performance depending on the requests arrivals characteristics and the time available to obtain a memory allocation.

Federated learning (FL) allows model training from local data collected by edge/mobile devices while preserving data privacy, which has wide applicability to image and vision applications. A challenge is that client devices in FL usually have much more limited computation and communication resources compared to servers in a data center. To overcome this challenge, we propose PruneFL -a novel FL approach with adaptive and distributed parameter pruning, which adapts the model size during FL to reduce both communication and computation overhead and minimize the overall training time, while maintaining a similar accuracy as the original model. PruneFL includes initial pruning at a selected client and further pruning as part of the FL process. The model size is adapted during this process, which includes maximizing the approximate empirical risk reduction divided by the time of one FL round. Our experiments with various datasets on edge devices (e.g., Raspberry Pi) show that: 1) we significantly reduce the training time compared to conventional FL and various other pruning-based methods and 2) the pruned model with automatically determined size converges to an accuracy that is very similar to the original model, and it is also a lottery ticket of the original model.

Data centers are increasingly using high-speed circuit switches to cope with the growing demand and reduce operational costs. One of the fundamental tasks of circuit switches is to compute a sparse collection of switching configurations to support a traffic demand matrix. Such a problem has been addressed in the literature with variations of the approach proposed by Birkhoff in 1946 to decompose a doubly stochastic matrix exactly. However, the existing methods are heuristic and do not have theoretical guarantees on how well a collection of switching configurations (i.e., permutations) can approximate a traffic matrix (i.e., a scaled doubly stochastic matrix). In this paper, we revisit Birkhoff’s approach and make three contributions. First, we establish the first theoretical bound on the sparsity of Birkhoff’s algorithm (i.e., the number of switching configurations necessary to approximate a traffic matrix). In particular, we show that by using a subset of the admissible permutation matrices, Birkhoff’s algorithm obtains an$\epsilon $-approximate decomposition with at most$O(\log (1 / \epsilon))$permutations. Second, we propose a new algorithm,Birkhoff+, which combines the wealth of Frank-Wolfe with Birkhoff’s approach to obtain sparse decompositions in a fast manner. And third, we evaluate the performance of the proposed algorithm numerically and study how this affects the performance of a circuit switch. Our results show thatBirkhoff+is superior to previous algorithms in terms of throughput, running time, and number of switching configurations.

This paper studies Online Convex Optimization (OCO) problems where the constraints have additive perturbations that (i) vary over time and (ii) are not known at the time to make a decision. Perturbations may not be i.i.d. generated and can be used, for example, to model a time-varying budget or time-varying requests in resource allocation problems. Our goal is to design a policy that obtains sublinear regret and satisfies the constraints in the long-term. To this end, we present an online primal-dual proximal gradient algorithm that has $O(T^\epsilon \vee T^{1-\epsilon})$ regret and $O(T^\epsilon)$ constraint violation, where $\epsilon \in [0,1)$ is a parameter in the learning rate. The proposed algorithm obtains optimal rates when $\epsilon = 1/2$, and can compare against a stronger comparator (the set of fixed decisions in hindsight) than previous work.

We develop an edge-assisted object recognition system with the aim of studying the system-level trade-offs between end-to-end latency and object recognition accuracy. We focus on developing techniques that optimize the transmission delay of the system and demonstrate the effect of image encoding rate and neural network size on these two performance metrics. We explore optimal trade-offs between these metrics by measuring the performance of our real time object recognition application. Our measurements reveal hitherto unknown parameter effects and sharp trade-offs, hence paving the road for optimizing this key service. Finally, we formulate two optimization problems using our measurement-based models and following a Pareto analysis we find that careful tuning of the system operation yields at least 33% better performance for real time conditions, over the standard transmission method.

We study the problem of in-network execution of data analytic services using multi-grade VNF chains. The nodes host VNFs offering different and possibly time-varying gains for each stage of the chain, and our goal is to maximize the analytics performance while minimizing the data transfer and processing costs. The VNFs' performance is revealed only after their execution, since it is data-dependent or controlled by third-parties, while the service requests and network costs might also vary with time. We devise an operation algorithm that learns, on the fly, the optimal routing policy and the composition and length of each chain. Our algorithm combines a lightweight sampling technique and a Lagrange-based primal-dual iteration, allowing it to be scalable and attain provable optimality guarantees. We demonstrate the performance of the proposed algorithm using a video analytics service, and explore how it is affected by different system parameters. Our model and optimization framework is readily extensible to different types of networks and services.

We propose an online policy that schedules the transmission and processing of data analytic tasks in an Internet of Things (IoT) network. The tasks are executed with different precision at the (possibly heterogeneous) nodes; the network is subject to link bandwidth and node processing capacity changes; and the task requests vary following unknown statistics. For this general IoT scenario, we formulate a resource allocation problem towards maximizing the aggregate tasks precision, and design a dynamic solution policy by combining the FrankWolfe and dual subgradient algorithms. Our policy (FWDS) is guaranteed to converge within bounded distance from the optimal solution, while ensuring interference-free transmissions, and being oblivious to network and/or traffic load changes. We use a wireless testbed and a state-of-the-art face recognition application to implement FWDS and compare it with static or dynamic (maxweight-type) competitor policies. Our findings verify that FWDS pushes the envelope of network control algorithms by handling time-varying objectives and possibly non-i.i.d. network/load statistics, with smaller complexity than its competitors.

Network densification over space and spectrum is expected to be key to enabling the requirements of next generation mobile systems. The pitfall is that radio resource allocation becomes substantially more complex. In this paper, we propose LaSR, a practical multi-connectivity scheduler for OFDMA-based multi-RAT systems. LaSR makes optimal discrete control actions by solving a sequence of simple optimization problems that do not require prior information of traffic patterns. In marked contrast to previous work, the flexibility of our approach allows us to construct scheduling policies that achieve a good balance between system cost and utility satisfaction, while jointly operate across heterogeneous RATs, accommodate real-system requirements, and guarantee system stability. Examples of system requirements considered in this paper include (but are not limited to): constraints on how scheduling data can be encoded onto signaling protocols (e.g., LTE's DCI), delays when turning on/off radio units, or on/off cycles when using unlicensed spectrum. We evaluate our scheduler via a thorough simulation campaign in a variety of scenarios with e.g., mobile users, RATs using unlicensed spectrum (using a duty cycle access mechanism), imperfect queue state information, and constrained signaling protocol.

This paper studies the problem of allocating band-width and computation resources to data analytics tasks in Internet of Things (IoT) networks. IoT nodes are powered by batteries, can process (some of) the data locally, and the quality grade or performance of how data analytics tasks are carried out depends on where these are executed. The goal is to design a resource allocation algorithm that jointly maximizes the network lifetime and the performance of the data analytics tasks subject to energy constraints. This joint maximization problem is challenging with coupled resource constraints that induce non-convexity. We first show that the problem can be mapped to an equivalent convex problem, and then propose an online algorithm that provably solves the problem and does not require any a priori knowledge of the time-varying wireless link capacities and data analytics arrival process statistics. The algorithm's optimality properties are derived using an analysis which, to the best of our knowledge, proves for the first time the convergence of the dual subgradient method with time-varying sets. Our simulations seeded by real IoT device energy measurements, show that the network connectivity plays a crucial role in network lifetime maximization, that the algorithm can obtain both maximum network lifetime and maximum data analytics performance in addition to maximizing the joint objective, and that the algorithm increases the network lifetime by approximately 50% compared to an algorithm that minimizes the total energy consumption.

In this paper, we study the problem of how to allocate bandwidth and computation resources to deliver data analytics services at the edge. The types of services we envision consist of a chain of tasks that must be carried out sequentially, and where the number of tasks executed in the chain determines the grade in which a service is delivered. An example of such type of service is video analytics where different deep-learning algorithms are combined to provide a more accurate description of a scene. The contributions of the paper are to formulate the static resource allocation problem as a linear program, to discuss the challenges of static formulations in dynamic settings, and to propose a control-type formulation that uses approximate system dynamics and time-varying cost functions. The work also highlights the need for policies that can operate the network and learn its characteristics simultaneously.

Future mobile networks will exploit unlicensed spectrum to boost capacity and meet growing user demands cost-effectively. The 3rdGeneration Partnership Project (3GPP) has recently defined a License Assisted Access (LAA) scheme to enable global Unlicensed LTE (U-LTE) deployment, aiming at 1) ensuring fair coexistence with incumbent WiFi networks, i.e., impacting on their performance no more than another WiFi device; and 2) achieving superior airtime efficiency as compared with WiFi. We show the standardized LAA fails to simultaneously fulfill these objectives, and design an alternative orthogonal (collision-free) listen-before-talk coexistence paradigm that provides a substantial improvement in performance, yet imposes no penalty on existing WiFi networks. We derive two optimal transmission policies, ORLA and OLAA, that maximize LAA throughput in both asynchronous and synchronous (i.e., with alignment to licensed anchor frame boundaries) modes of operation, respectively. We present a comprehensive evaluation through which we demonstrate that, when aggregating packets, IEEE 802.11ac WiFi can be more efficient than LAA, whereas our proposals attains 100% higher throughput, without harming WiFi. We further show that long U-LTE frames incur up to 92% throughput losses on WiFi when using 3GPP LAA, whilst ORLA/OLAA sustain >200% gains at no cost, even in the presence of non-saturated WiFi and/or in multi-rate scenarios.

We investigate the connections between max-weight approaches and dual subgradient methods for convex optimization. We find that strong connections exist, and we establish a clean, unifying theoretical framework that includes both max-weight and dual subgradient approaches as special cases. Our analysis uses only elementary methods and is not asymptotic in nature. It also allows us to establish an explicit and direct connection between discrete queue occupancies and Lagrange multipliers.