There is a recently discovered and intriguing phenomenon called Neural Collapse: at the terminal phase of training a deep neural network for classification, the within-class penultimate feature means and the associated classifier vectors of all flat classes collapse to the vertices of a simplex Equiangular Tight Frame (ETF). Recent work has tried to exploit this phenomenon by fixing the related classifier weights to a pre-computed ETF to induce neural collapse and maximize the separation of the learned features when training with imbalanced data. In this work, we propose to fix the linear classifier of a deep neural network to a Hierarchy-Aware Frame (HAFrame), instead of an ETF, and use a cosine similarity-based auxiliary loss to learn hierarchy-aware penultimate features that collapse to the HAFrame. We demonstrate that our approach reduces the mistake severity of the model's predictions while maintaining its top-1 accuracy on several datasets of varying scales with hierarchies of heights ranging from 3 to 12. Code: https://github.com/ltong1130ztr/HAFrame.

Recent developments of deep learning methods have demonstrated their feasibility in liver malignancy diagnosis using ultrasound (US) images. However, most of these methods require manual selection and annotation of US images by radiologists, which limit their practical application. On the other hand, US videos provide more comprehensive morphological information about liver masses and their relationships with surrounding structures than US images, potentially leading to a more accurate diagnosis. Here, we developed a fully automated artificial intelligence (AI) pipeline to imitate the workflow of radiologists for detecting liver masses and diagnosing liver malignancy. In this pipeline, we designed an automated mass-guided strategy that used segmentation information to direct diagnostic models to focus on liver masses, thus increasing diagnostic accuracy. The diagnostic models based on US videos utilized bi-directional convolutional long short-term memory modules with an attention-boosted module to learn and fuse spatiotemporal information from consecutive video frames. Using a large-scale dataset of 50 063 US images and video frames from 11 468 patients, we developed and tested the AI pipeline and investigated its applications. A dataset of annotated US images is available at https://doi.org/10.5281/zenodo.7272660.

The integration of reinforcement learning with fuzzy logic can be effective in compensating the external disturbance and complex dynamic while designing the control strategy for vehicle suspension. The main contribution of this paper is that a learning-based adaptive vibration control strategy is proposed for semi-active suspension system, which combines the fuzzy logic with the reward function of reinforcement learning to improve the robustness and feasibility of the vibration control strategy. What’s more, an improved proximal policy optimization algorithm combined with fuzzy logic is proposed for realizing the trial-and-error reinforcement learning. Specially, the reward function with fuzzy logic is formulated to meet the requirements of suspension performance under different road conditions, in which the fuzzy logic is designed to fuzzily the process the collected road information, real-time update the weight matrix coefficients, and adjust the optimization objectives adaptively. Finally, numerical simulation results are given to prove the effectiveness of the proposed vibration control strategy.

Many organism behaviors are innate or instinctual and have been "hard-coded" through evolution. Current approaches to understanding these behaviors model evolution as an optimization problem in which the traits of organisms are assumed to optimize an objective function representing evolutionary fitness. Here, we use a mechanistic birth-death dynamics approach to study the evolution of innate behavioral strategies in a simulated population of organisms. In particular, we performed agent-based stochastic simulations and mean-field analyses of organisms exploring random environments and competing with each other to find locations with plentiful resources. We find that when organism density is low, the mean-field model allows us to derive an effective objective function, predicting how the most competitive phenotypes depend on the exploration-exploitation trade-off between the scarcity of high-resource sites and the increase in birth rate those sites offer organisms. However, increasing organism density alters the most competitive behavioral strategies and precludes the derivation of a well-defined objective function. Moreover, there exists a range of densities for which the coexistence of many phenotypes persists for evolutionarily long times.

We present a bottom-up approach to hierarchical classification based on posteriors conditioned with logits. Beginning with the output logits for a set of terminal labels from a base classifier, an initial hypothesis is repeatedly generalized (softened) to a weaker label until a particular confidence measure is achieved. As conditioning the probabilistic model with the full set of terminal logits quickly becomes intractable for large label sets, we propose an alternative approach employing “generalized logits” spanning relevant hypotheses within the label hierarchy. Experimental results are compared with related methods on multiple datasets and base classifiers. The proposed approach provides an efficient and effective hierarchical classification framework with monotonic, non-decreasing inference behavior.

MOTIVATION: Repeats detection problems are traditionally formulated as string matching or signal processing problems. They cannot readily handle gaps between repeat units and are incapable of detecting repeat patterns shared by multiple sequences. This study detects short adjacent repeats with interunit insertions from multiple sequences. For biological sequences, such studies can shed light on molecular structure, biological function and evolution. RESULTS: The task of detecting short adjacent repeats is formulated as a statistical inference problem by using a probabilistic generative model. An Markov chain Monte Carlo algorithm is proposed to infer the parameters in a de novo fashion. Its applications on synthetic and real biological data show that the new method not only has a competitive edge over existing methods, but also can provide a way to study the structure and the evolution of repeat-containing genes. AVAILABILITY: The related C++ source code and datasets are available at http://ihome.cuhk.edu.hk/%7Eb118998/share/BASARD.zip. CONTACT: [email protected]

Fluorescence in situ hybridization (FISH) is a powerful technique that localizes specific DNA sequences on chromosomes for use in physical and genetic maps assembling, genetic counselling, species identification, etc. Highly repetitive sequences are considered to be suitable FISH probes that can avoid many potential problems of using unique sequences as FISH probes. The distinct chromosomal distributions of these highly repetitive sequences are also ideal for labelling purposes such as karyotyping. In this paper, we present an automatic computational procedure for searching highly repetitive sequences from a whole genome as FISH probes, as well as an experimental protocol to use them in FISH analysis. We successfully applied the method on the newly released genome of Brachypodium distachyon (Brachypodium) and produced satisfactory results of FISH experiment.

Relay selection is a challenging issue in cooperative communication networks. In particular, in a multi-user multi- concurrent-transmission network, effects of the resource sharing of a relay by its served users on the relay selection strategy need to be taken into account. There is little research on this investigation in literature. In this paper, we propose a novel efficient and fair relay selection strategy, in which individual user's QoS requirement is inherently taken into account. In terms of efficiency, the proposed strategy is of low computational complexity. In terms of fairness, the system resource is utilized fairly among all relaying nodes so that the system has the best potential to accommodate as many users as possible. Simulation results indicate that the proposed selection strategy outperforms many existing ones, therefore justifying our arguments.