Non-intrusive load monitoring (NILM), also known as energy disaggregation, is a blind source separation problem where a household’s aggregate electricity consumption is broken down into electricity usages of individual appliances. In this way, the cost and trouble of installing many measurement devices over numerous household appliances can be avoided, and only one device needs to be installed. The problem has been well-known since Hart’s seminal paper in 1992, and recently significant performance improvements have been achieved by adopting deep networks. In this work, we focus on the idea that appliances have on/off states, and develop a deep network for further performance improvements. Specifically, we propose a subtask gated network that combines the main regression network with an on/off classification subtask network. Unlike typical multitask learning algorithms where multiple tasks simply share the network parameters to take advantage of the relevance among tasks, the subtask gated network multiply the main network’s regression output with the subtask’s classification probability. When standby-power is additionally learned, the proposed solution surpasses the state-of-the-art performance for most of the benchmark cases. The subtask gated network can be very effective for any problem that inherently has on/off states.

In designing new service plans, network service providers need to understand how consumption of voice or data service will change in response to pricing signals. It is difficult to acquire such information from customer usage data because voice minutes and data bandwidth are typically sold in the form of large quotas. We address this issue by studying how end-users consume their quotas, both in a prepaid setting (where users pay in advance and refill as needed) and a postpaid setting (where users pay each month for a fixed amount of quota). Our presentation has three main parts. In the first we present data on quota usage for prepaid voice/text services and show that users reduce their voice usage when their balances become low. Moreover, when balances are low there is a tendency to shift from voice to SMS. In the second part, we provide descriptive models of both prepaid and postpaid services. The main feature of these models is that there is a background level of potential demand and the rate at which this potential demand is realized depends on the amount of quota balance available. In the third part, we propose utility maximizing models that can account for this type of behavior. In the prepaid case the main feature of the model is a discount function that represents the perceived cost to the user of a quota refill that will occur sometime in the future. In the postpaid case, where the end-user is attempting to get the maximum amount of utility from his monthly quota, we present a dynamic programming formulation in which utility functions are time varying and not known to the user in advance.

We propose an adaptive equalizer for coded communication systems, which adjusts filter taps using the gradient descent algorithm under the greatest constraint satisfaction (GCS) criterion. In particular, we derive the GCS equalizer for a low-density parity-check (LDPC)-coded multiple-input multiple-output (MIMO) system. We demonstrate by simulations that the GCS equalizer significantly outperforms the minimum mean square error (MMSE) equalizer under binary phase-shift keying (BPSK) modulation, and performs similar under quadrature phase-shift keying (QPSK) modulation.