-
ResOT: Resource-Efficient Oblique Trees for Neural Signal Classification
Authors:
Bingzhao Zhu,
Masoud Farivar,
Mahsa Shoaran
Abstract:
Classifiers that can be implemented on chip with minimal computational and memory resources are essential for edge computing in emerging applications such as medical and IoT devices. This paper introduces a machine learning model based on oblique decision trees to enable resource-efficient classification on a neural implant. By integrating model compression with probabilistic routing and implement…
▽ More
Classifiers that can be implemented on chip with minimal computational and memory resources are essential for edge computing in emerging applications such as medical and IoT devices. This paper introduces a machine learning model based on oblique decision trees to enable resource-efficient classification on a neural implant. By integrating model compression with probabilistic routing and implementing cost-aware learning, our proposed model could significantly reduce the memory and hardware cost compared to state-of-the-art models, while maintaining the classification accuracy. We trained the resource-efficient oblique tree with power-efficient regularization (ResOT-PE) on three neural classification tasks to evaluate the performance, memory, and hardware requirements. On seizure detection task, we were able to reduce the model size by 3.4X and the feature extraction cost by 14.6X compared to the ensemble of boosted trees, using the intracranial EEG from 10 epilepsy patients. In a second experiment, we tested the ResOT-PE model on tremor detection for Parkinson's disease, using the local field potentials from 12 patients implanted with a deep-brain stimulation (DBS) device. We achieved a comparable classification performance as the state-of-the-art boosted tree ensemble, while reducing the model size and feature extraction cost by 10.6X and 6.8X, respectively. We also tested on a 6-class finger movement detection task using ECoG recordings from 9 subjects, reducing the model size by 17.6X and feature computation cost by 5.1X. The proposed model can enable a low-power and memory-efficient implementation of classifiers for real-time neurological disease detection and motor decoding.
△ Less
Submitted 14 June, 2020;
originally announced June 2020.
-
Optimal Inverter VAR Control in Distribution Systems with High PV Penetration
Authors:
Masoud Farivar,
Russell Neal,
Christopher Clarke,
Steven Low
Abstract:
The intent of the study detailed in this paper is to demonstrate the benefits of inverter var control on a fast timescale to mitigate rapid and large voltage fluctuations due to the high penetration of photovoltaic generation and the resulting reverse power flow. Our approach is to formulate the volt/var control as a radial optimal power flow (OPF) problem to minimize line losses and energy consum…
▽ More
The intent of the study detailed in this paper is to demonstrate the benefits of inverter var control on a fast timescale to mitigate rapid and large voltage fluctuations due to the high penetration of photovoltaic generation and the resulting reverse power flow. Our approach is to formulate the volt/var control as a radial optimal power flow (OPF) problem to minimize line losses and energy consumption, subject to constraints on voltage magnitudes. An efficient solution to the radial OPF problem is presented and used to study the structure of optimal inverter var injection and the net benefits, taking into account the additional cost of inverter losses when operating at non-unity power factor. This paper will illustrate how, depending on the circuit topology and its loading condition, the inverter's optimal reactive power injection is not necessarily monotone with respect to their real power output. The results are demonstrated on a distribution feeder on the Southern California Edison system that has a very light load and a 5 MW photovoltaic (PV) system installed away from the substation.
△ Less
Submitted 23 December, 2011;
originally announced December 2011.
-
Two Dimensional Connectivity for Vehicular Ad-Hoc Networks
Authors:
Masoud Farivar,
Behzad Mehrdad,
Farid Ashtiani
Abstract:
In this paper, we focus on two-dimensional connectivity in sparse vehicular ad hoc networks (VANETs). In this respect, we find thresholds for the arrival rates of vehicles at entrances of a block of streets such that the connectivity is guaranteed for any desired probability. To this end, we exploit a mobility model recently proposed for sparse VANETs, based on BCMP open queuing networks and sol…
▽ More
In this paper, we focus on two-dimensional connectivity in sparse vehicular ad hoc networks (VANETs). In this respect, we find thresholds for the arrival rates of vehicles at entrances of a block of streets such that the connectivity is guaranteed for any desired probability. To this end, we exploit a mobility model recently proposed for sparse VANETs, based on BCMP open queuing networks and solve the related traffic equations to find the traffic characteristics of each street and use the results to compute the exact probability of connectivity along these streets. Then, we use the results from percolation theory and the proposed fast algorithms for evaluation of bond percolation problem in a random graph corresponding to the block of the streets. We then find sufficiently accurate two dimensional connectivity-related parameters, such as the average number of intersections connected to each other and the size of the largest set of inter-connected intersections. We have also proposed lower bounds for the case of heterogeneous network with two transmission ranges. In the last part of the paper, we apply our method to several numerical examples and confirm our results by simulations.
△ Less
Submitted 9 December, 2008;
originally announced December 2008.
-
Approximate Sparse Decomposition Based on Smoothed L0-Norm
Authors:
Hamed Firouzi,
Masoud Farivar,
Massoud Babaie-Zadeh,
Christian Jutten
Abstract:
In this paper, we propose a method to address the problem of source estimation for Sparse Component Analysis (SCA) in the presence of additive noise. Our method is a generalization of a recently proposed method (SL0), which has the advantage of directly minimizing the L0-norm instead of L1-norm, while being very fast. SL0 is based on minimization of the smoothed L0-norm subject to As=x. In order…
▽ More
In this paper, we propose a method to address the problem of source estimation for Sparse Component Analysis (SCA) in the presence of additive noise. Our method is a generalization of a recently proposed method (SL0), which has the advantage of directly minimizing the L0-norm instead of L1-norm, while being very fast. SL0 is based on minimization of the smoothed L0-norm subject to As=x. In order to better estimate the source vector for noisy mixtures, we suggest then to remove the constraint As=x, by relaxing exact equality to an approximation (we call our method Smoothed L0-norm Denoising or SL0DN). The final result can then be obtained by minimization of a proper linear combination of the smoothed L0-norm and a cost function for the approximation. Experimental results emphasize on the significant enhancement of the modified method in noisy cases.
△ Less
Submitted 18 November, 2008;
originally announced November 2008.
