Source author record

Junjun Liang

Junjun Liang appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

cond-mat.quant-gas Computation and Language quant-ph

Catalog footprint

What is connected

4works

3topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2020arXiv

Dice Loss for Data-imbalanced NLP Tasks

Many NLP tasks such as tagging and machine reading comprehension are faced with the severe data imbalance issue: negative examples significantly outnumber positive examples, and the huge number of background examples (or easy-negative examples) overwhelms the training. The most commonly used cross entropy (CE) criteria is actually an accuracy-oriented objective, and thus creates a discrepancy between training and test: at training time, each training instance contributes equally to the objective function, while at test time F1 score concerns more about positive examples. In this paper, we propose to use dice loss in replacement of the standard cross-entropy objective for data-imbalanced NLP tasks. Dice loss is based on the Sorensen-Dice coefficient or Tversky index, which attaches similar importance to false positives and false negatives, and is more immune to the data-imbalance issue. To further alleviate the dominating influence from easy-negative examples in training, we propose to associate training examples with dynamically adjusted weights to deemphasize easy-negative examples.Theoretical analysis shows that this strategy narrows down the gap between the F1 score in evaluation and the dice loss in training. With the proposed training objective, we observe significant performance boost on a wide range of data imbalanced NLP tasks. Notably, we are able to achieve SOTA results on CTB5, CTB6 and UD1.4 for the part of speech tagging task; SOTA results on CoNLL03, OntoNotes5.0, MSRA and OntoNotes4.0 for the named entity recognition task; along with competitive results on the tasks of machine reading comprehension and paraphrase identification.

preprint2015arXiv

Spin-orbit coupled repulsive Fermi atoms in a one-dimensional optical lattice

Motivated by recent experimental development, we investigate spin-orbit coupled repulsive Fermi atoms in a one-dimensional optical lattice. Using the density-matrix renormalization group method, we calculate momentum distribution function, gap, and spin-correlation function to reveal rich ground-state properties. We find that spin-orbit coupling (SOC) can generate unconventional momentum distribution, which depends crucially on the filling. We call the corresponding phase with zero gap the SOC-induced metallic phase. We also show that SOC can drive the system from the antiferromagnetic to ferromagnetic Mott insulators with spin rotating. As a result, a second-order quantum phase transition between the spin-rotating ferromagnetic Mott insulator and the SOC-induced metallic phase is predicted at the strong SOC. Here the spin rotating means that the spin orientations of the nearest-neighbor sites are not parallel or antiparallel, i.e., they have an intersection angle $θ\in (0,π)$. Finally, we show that the momentum $k_{\mathrm{peak}}$, at which peak of the spin-structure factor appears, can also be affected dramatically by SOC. The analytical expression of this momentum with respect to the SOC strength is also derived. It suggests that the predicted spin-rotating ferromagnetic ($k_{\mathrm{peak}% }<π/2$) and antiferromagnetic ($π/2<k_{\mathrm{peak}}<π$) correlations can be detected experimentally by measuring the SOC-dependent spin-structure factor via the time-of-flight imaging.

preprint2015arXiv

Unconventional pairings of spin-orbit coupled attractive degenerate Fermi gas in a one dimensional optical lattice

Understanding novel pairings in attractive degenerate Fermi gases is crucial for exploring rich superfluid physics. In this report, we reveal unconventional pairings induced by spin-orbit coupling (SOC) in a one-dimensional optical lattice, using a state-of-the-art density-matrix renormalization group method. When both bands are partially occupied, we find a strong competition between the interband Fulde-Ferrell-Larkin-Ovchinnikov (FFLO) and intraband Bardeen-Cooper-Schrieffer (BCS) pairings. In particular, for the weak and moderate SOC strengths, these two pairings can coexist, giving rise to a new phase called the FFLO-BCS phase, which exhibits a unique three-peak structure in pairing momentum distribution. For the strong SOC strength, the intraband BCS pairing always dominates in the whole parameter regime, including the half filling. We figure out the whole phase diagrams as functions of filling factor, SOC strength, and Zeeman field. Our results are qualitatively different from recent mean-field predictions. Finally, we address that our predictions could be observed in a weaker trapped potential.

preprint2012arXiv

Pairing imbalance in BCS-BEC crossover of inhomogeneous three-component Fermi-gas in two dimensions

We in this paper investigate the phase diagram associated with the BCS-BEC crossover of a three-component ultracold superfluid-Fermi-gas of different chemical-potentials and equal masses in two dimensions. The gap order parameter and number densities are found analytically by using the functional path-integral method. The balance of paring will be broken in the free space due to the unequal chemical-potentials. We obtain the same particle number-density and condensed fraction in the BCS superfluid phase as that in a recent paper (Phys. Rev. A 83, 033630), while the Sarma phase of coexistence of normal and superfluid Fermi gases is the characteristics of inhomogeneous system. The minimum ratio of BCS superfluid phase becomes 1/3 in the BCS limit corresponding to the zero-ratio in the two-component system in which the critical point of phase separation is εB/εF = 2 but becomes 3 in the three-component case.