-
Modeling of learning curves with applications to pos tagging
Authors:
Manuel Vilares Ferro,
Victor M. Darriba Bilbao,
Francisco J. Ribadas Pena
Abstract:
An algorithm to estimate the evolution of learning curves on the whole of a training data base, based on the results obtained from a portion and using a functional strategy, is introduced. We approximate iteratively the sought value at the desired time, independently of the learning technique used and once a point in the process, called prediction level, has been passed. The proposal proves to be…
▽ More
An algorithm to estimate the evolution of learning curves on the whole of a training data base, based on the results obtained from a portion and using a functional strategy, is introduced. We approximate iteratively the sought value at the desired time, independently of the learning technique used and once a point in the process, called prediction level, has been passed. The proposal proves to be formally correct with respect to our working hypotheses and includes a reliable proximity condition. This allows the user to fix a convergence threshold with respect to the accuracy finally achievable, which extends the concept of stop** criterion and seems to be effective even in the presence of distorting observations.
Our aim is to evaluate the training effort, supporting decision making in order to reduce the need for both human and computational resources during the learning process. The proposal is of interest in at least three operational procedures. The first is the anticipation of accuracy gain, with the purpose of measuring how much work is needed to achieve a certain degree of performance. The second relates the comparison of efficiency between systems at training time, with the objective of completing this task only for the one that best suits our requirements. The prediction of accuracy is also a valuable item of information for customizing systems, since we can estimate in advance the impact of settings on both the performance and the development costs. Using the generation of part-of-speech taggers as an example application, the experimental results are consistent with our expectations.
△ Less
Submitted 4 February, 2024;
originally announced February 2024.
-
Early stop** by correlating online indicators in neural networks
Authors:
Manuel Vilares Ferro,
Yerai Doval Mosquera,
Francisco J. Ribadas Pena,
Victor M. Darriba Bilbao
Abstract:
In order to minimize the generalization error in neural networks, a novel technique to identify overfitting phenomena when training the learner is formally introduced. This enables support of a reliable and trustworthy early stop** condition, thus improving the predictive power of that type of modeling. Our proposal exploits the correlation over time in a collection of online indicators, namely…
▽ More
In order to minimize the generalization error in neural networks, a novel technique to identify overfitting phenomena when training the learner is formally introduced. This enables support of a reliable and trustworthy early stop** condition, thus improving the predictive power of that type of modeling. Our proposal exploits the correlation over time in a collection of online indicators, namely characteristic functions for indicating if a set of hypotheses are met, associated with a range of independent stop** conditions built from a canary judgment to evaluate the presence of overfitting. That way, we provide a formal basis for decision making in terms of interrupting the learning process.
As opposed to previous approaches focused on a single criterion, we take advantage of subsidiarities between independent assessments, thus seeking both a wider operating range and greater diagnostic reliability. With a view to illustrating the effectiveness of the halting condition described, we choose to work in the sphere of natural language processing, an operational continuum increasingly based on machine learning. As a case study, we focus on parser generation, one of the most demanding and complex tasks in the domain. The selection of cross-validation as a canary function enables an actual comparison with the most representative early stop** conditions based on overfitting identification, pointing to a promising start toward an optimal bias and variance control.
△ Less
Submitted 4 February, 2024;
originally announced February 2024.
-
Measurement of the ionization produced by sub-keV silicon nuclear recoils in a CCD dark matter detector
Authors:
A. E. Chavarria,
J. I. Collar,
J. R. Peña,
P. Privitera,
A. E. Robinson,
B. Scholz,
C. Sengul,
J. Zhou,
J. Estrada,
F. Izraelevitch,
J. Tiffenberg,
J. R. T. de Mello Neto,
D. Torres Machado
Abstract:
We report a measurement of the ionization efficiency of silicon nuclei recoiling with sub-keV kinetic energy in the bulk silicon of a charge-coupled device (CCD). Nuclear recoils are produced by low-energy neutrons ($<$24 keV) from a $^{124}$Sb-$^{9}$Be photoneutron source, and their ionization signal is measured down to 60 eV electron equivalent. This energy range, previously unexplored, is relev…
▽ More
We report a measurement of the ionization efficiency of silicon nuclei recoiling with sub-keV kinetic energy in the bulk silicon of a charge-coupled device (CCD). Nuclear recoils are produced by low-energy neutrons ($<$24 keV) from a $^{124}$Sb-$^{9}$Be photoneutron source, and their ionization signal is measured down to 60 eV electron equivalent. This energy range, previously unexplored, is relevant for the detection of low-mass dark matter particles. The measured efficiency is found to deviate from the extrapolation to low energies of the Lindhard model. This measurement also demonstrates the sensitivity to nuclear recoils of CCDs employed by DAMIC, a dark matter direct detection experiment located in the SNOLAB underground laboratory.
△ Less
Submitted 9 November, 2016; v1 submitted 2 August, 2016;
originally announced August 2016.
-
Search for low-mass WIMPs in a 0.6 kg day exposure of the DAMIC experiment at SNOLAB
Authors:
A. Aguilar-Arevalo,
D. Amidei,
X. Bertou,
M. Butner,
G. Cancelo,
A. Castañeda Vázquez,
B. A. Cervantes Vergara,
A. E. Chavarria,
C. R. Chavez,
J. R. T. de Mello Neto,
J. C. D'Olivo,
J. Estrada,
G. Fernandez Moroni,
R. Gaïor,
Y. Guandincerri,
K. P. Hernández Torres,
F. Izraelevitch,
A. Kavner,
B. Kilminster,
I. Lawson,
A. Letessier-Selvon,
J. Liao,
J. Molina,
J. R. Peña,
P. Privitera
, et al. (13 additional authors not shown)
Abstract:
We present results of a dark matter search performed with a 0.6 kg day exposure of the DAMIC experiment at the SNOLAB underground laboratory. We measure the energy spectrum of ionization events in the bulk silicon of charge-coupled devices down to a signal of 60 eV electron equivalent. The data are consistent with radiogenic backgrounds, and constraints on the spin-independent WIMP-nucleon elastic…
▽ More
We present results of a dark matter search performed with a 0.6 kg day exposure of the DAMIC experiment at the SNOLAB underground laboratory. We measure the energy spectrum of ionization events in the bulk silicon of charge-coupled devices down to a signal of 60 eV electron equivalent. The data are consistent with radiogenic backgrounds, and constraints on the spin-independent WIMP-nucleon elastic-scattering cross section are accordingly placed. A region of parameter space relevant to the potential signal from the CDMS-II Si experiment is excluded using the same target for the first time. This result obtained with a limited exposure demonstrates the potential to explore the low-mass WIMP region (<10 GeV/$c^{2}$) of the upcoming DAMIC100, a 100 g detector currently being installed in SNOLAB.
△ Less
Submitted 9 November, 2016; v1 submitted 25 July, 2016;
originally announced July 2016.
