Image classification for real-world applications often involves complicated data distributions such as fine-grained and long-tailed. To address the two challenging issues simultaneously, we propose a new regularization technique that yields an adversarial loss to strengthen the model learning. Specifically, for each training batch, we construct an adaptive batch prediction (ABP) matrix and establish its corresponding adaptive batch confusion norm (ABC-Norm). The ABP matrix is a composition of two parts, including an adaptive component to class-wise encode the imbalanced data distribution, and the other component to batch-wise assess the softmax predictions. The ABC-Norm leads to a norm-based regularization loss, which can be theoretically shown to be an upper bound for an objective function closely related to rank minimization. By coupling with the conventional cross-entropy loss, the ABC-Norm regularization could introduce adaptive classification confusion and thus trigger adversarial learning to improve the effectiveness of model learning. Different from most of state-of-the-art techniques in solving either fine-grained or long-tailed problems, our method is characterized with its simple and efficient design, and most distinctively, provides a unified solution. In the experiments, we compare ABC-Norm with relevant techniques and demonstrate its efficacy on several benchmark datasets, including (CUB-LT, iNaturalist2018); (CUB, CAR, AIR); and (ImageNet-LT), which respectively correspond to the real-world, fine-grained, and long-tailed scenarios.

Two types of randomness are associated with a mixed quantum state: the uncertainty in the probability coefficients of the constituent pure states and the uncertainty in the value of each observable captured by the Born’s rule probabilities. Entropy is a quantification of randomness, and we propose a spin-entropy for the observables of spin pure states based on the phase space of a spin as described by the geometric quantization method, and we also expand it to mixed quantum states. This proposed entropy overcomes the limitations of previously-proposed entropies such as von Neumann entropy which only quantifies the randomness of specifying the quantum state. As an example of a limitation, previously-proposed entropies are higher for Bell entangled spin states than for disentangled spin states, even though the spin observables are less constrained for a disentangled pair of spins than for an entangled pair. The proposed spin-entropy accurately quantifies the randomness of a quantum state, it never reaches zero value, and it is lower for entangled states than for disentangled states.

Quantum physics, despite its intrinsically probabilistic nature, lacks a definition of entropy fully accounting for the randomness of a quantum state. For example, von Neumann entropy quantifies only the incomplete specification of a quantum state and does not quantify the probabilistic distribution of its observables; it trivially vanishes for pure quantum states. We propose a quantum entropy that quantifies the randomness of a pure quantum state via a conjugate pair of observables/operators forming the quantum phase space. The entropy is dimensionless, it is a relativistic scalar, it is invariant under canonical transformations and under CPT transformations, and its minimum has been established by the entropic uncertainty principle. We expand the entropy to also include mixed states. We show that the entropy is monotonically increasing during a time evolution of coherent states under a Dirac Hamiltonian. However, in a mathematical scenario, when two fermions come closer to each other, each evolving as a coherent state, the total system’s entropy oscillates due to the increasing spatial entanglement. We hypothesize an entropy law governing physical systems whereby the entropy of a closed system never decreases, implying a time arrow for particle physics. We then explore the possibility that as the oscillations of the entropy must by the law be barred in quantum physics, potential entropy oscillations trigger annihilation and creation of particles.

Quantum physics through the lens of Bayesian statistics considers probability to be a degree of belief and subjective. A Bayesian derivation of the probability density function in phase space is presented. Then, a Kullback–Liebler divergence in phase space is introduced to define interference and entanglement. Comparisons between each of these two quantities and the entropy are made. A brief presentation of entanglement in phase space to the spin degree of freedom and an extension to mixed states completes the work.

Quantum physics through the lenses of Bayesian statistics consider probabilities to be degree of believes or knowledge and subjective. A Bayesian derivation of the probability density function in phase space is presented. The entropy proposed by Geiger and Kedem \cite{GeiKed2022a,GeiKed2022b} in phase space is expanded to define quantitatively interference and entanglement. Comparisons between each of these two quantities and the entropy are made. A brief presentation of entanglement in phase space to the spin degree of freedom and to mixed states completes the work.