Motivation Human genomic datasets often contain sensitive information that limits use and sharing of the data. In particular, simple anonymization strategies fail to provide sufficient level of protection for genomic data, because the data are inherently identifiable. Differentially private machine learning can help by guaranteeing that the published results do not leak too much information about any individual data point. Recent research has reached promising results on differentially private drug sensitivity prediction using gene expression data. Differentially private learning with genomic data is challenging because it is more difficult to guarantee privacy in high dimensions. Dimensionality reduction can help, but if the dimension reduction mapping is learned from the data, then it needs to be differentially private too, which can carry a significant privacy cost. Furthermore, the selection of any hyperparameters (such as the target dimensionality) needs to also avoid leaking private information. Results We study an approach that uses a large public dataset of similar type to learn a compact representation for differentially private learning. We compare three representation learning methods: variational autoencoders, principal component analysis and random projection. We solve two machine learning tasks on gene expression of cancer cell lines: cancer type classification, and drug sensitivity prediction. The experiments demonstrate significant benefit from all representation learning methods with variational autoencoders providing the most accurate predictions most often. Our results significantly improve over previous state-of-the-art in accuracy of differentially private drug sensitivity prediction. Availability and implementation Code used in the experiments is available at https://github.com/DPBayes/dp-representation-transfer.
In this paper, fuel consumption of a 5.7-ton municipal tractor in a wheel loader application is studied, and methods for improving the fuel efficiency are compared with each other. Experimental data from the baseline machine with load-sensing hydraulics has been gathered during a y-pattern cycle, and the data is inputted to an optimization function having realistic loss models for a hydraulic pump and diesel engine. Dynamic programming is used to analyze different system configurations in order to determine optimal control sequence for each system. Besides optimization of variable engine rotational speed on the baseline system during the working cycle (considering the point of operation), three hybrid supply systems are studied: 1) a hydraulic flywheel, 2) parallel supply pumps and 3) a throttled accumulator. These systems utilize a hydraulic accumulator as an energy source/sink alongside the diesel engine. The optimal sequence for charging and discharging of the accumulator is examined in order to minimize the fuel consumption of the machine. The idea is to use the lowest acceptable, constant engine rotational speed, to cut down the diesel losses. In addition, the study covers an analysis of adjusting the engine rotational speed for each point of operation also when the hybrid systems are considered. The results show that finding advantageous engine rotational speed for each loading condition can decrease the fuel consumption of the baseline machine around 14%, whereas hybridization of the supply system can further improve the result by a couple of percentage units. Hybrid systems also reduce engine’s maximum load by making it more uniform, which allegedly reduces emissions. The possibility of engine downsizing to further improve the fuel efficiency of hybrid systems is not considered, because the maximum engine power is usually determined by the hydrostatic transmission of a municipal tractor. However, the study assumes that actuators are controlled using traditional 4/3 proportional control valves; hence, there are still potential for greater fuel savings. For example, applying independent metering valves on the actuator control can further decrease the system losses.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.