Personalized Federated Learning with Semisupervised Distillation

Heterogeneous data and models pose critical challenges for federated learning. However, the traditional federated learning framework, which trains the global model by transferring model parameters, has major limitations; it requires that all participants have the same training model architectures, and the trained global model does not guarantee accurate projections for participants’ personal data. To solve this problem, we propose a new federal framework named personalized federated learning with semisupervised distillation (pFedSD), which ensures the privacy of the participants’ model architectures and improves the communication efficiency by transmitting the model’s predicted class distribution rather than model parameters. First, the server adopts the adaptive aggregation method to reduce the weight of low-quality model predictions for the model’s predicted class distributions uploaded by all clients, which helps to improve the quality of the aggregation of the prediction class distribution. Then, the server sends it back to the clients for local training to obtain the personalized model. We finally conducted experiments on different datasets (MNIST, FMNIST, and CIFAR10), and the results show that the model performance of pFedSD exceeds the latest federated distillation algorithms.

Vacas a matar. De la dicotomía soberana al umbral biopolítico en ficciones ganaderas del Cono Sur.

RESUMEN: Nudo de tensión entre vida biológica e historicidad, la animalidad recorre la serie de mataderos que la crítica argentina ha indagado como ficción soberana, pautada por el dualismo político. La polarización de 1840 plasmada en El matadero de Echeverría, con titubeos narrativos frente a voces y cuerpos animalizados del sector popular, retorna desde mediados del siglo XX en versiones de la dicotomía civilización-barbarie, que mantienen la distinción naturaleza-cultura y las coordenadas de historicidad antropocéntrica, como la crónica de Walsh que en 1967 enmienda el reparto clasista de Echeverría. El umbral crítico de la vida (Agamben) emerge en reescrituras recientes del matadero, donde el flujo del capital atraviesa no solo el cuerpo del trabajador sino lo viviente. Un cuento de Kohan (2009) y una novela de Maia (2013) descubren bajo la dicotomía la complejidad de pasajes entre animales y humanos, donde indagan las economías de relaciones entre vidas protegidas y abandonadas. Confrontando las conflictividades culturales argentina y brasileña, marcadas por el peso de economías agroexportadoras y el dualismo como control de mezclas, estos relatos de trabajadores del ganado inquieren lo difuso del borde vida-muerte tras dos siglos de progreso humanista en América Latina. ABSTRACT: Animality, a knot of tension between biological life and historicity, runs through the series of slaughterhouses that Argentine critics have investigated as sovereign fiction, guided by political dualism. The 1840 polarization embodied in Echeverría´s El matadero (The Slaughterhouse), with narrative hesitations about voices and animalized bodies of popular sector, has returned since the middle of the 20th century in versions of the civilization-barbarity dichotomy, which maintain the nature-culture distinction and the coordinates of anthropocentric historicity, such as Walsh's chronicle that in 1967 amends the class distribution of Echeverría. The critical threshold of life (Agamben) emerges in recent rewritings of the slaughterhouse, where the flow of capital passes through not only the body of the worker but also the living being. A story by Kohan (2009) and a novel by Maia (2013) discover under dichotomy the complexity of passages between animals and humans, where they investigate the economies of relationships between protected and abandoned lives. By confronting Argentine and Brazilian cultural conflicts, both marked by the weight of agro-export economies and dualism as control of mixtures, these stories of cattle workers inquire about the diffuseness of the life-death border after two centuries of humanist progress in Latin America.

A Learning Framework for Medical Image-Based Intelligent Diagnosis from Imbalanced Datasets

Medical image classification and diagnosis based on machine learning has made significant achievements and gradually penetrated the healthcare industry. However, medical data characteristics such as relatively small datasets for rare diseases or imbalance in class distribution for rare conditions significantly restrains their adoption and reuse. Imbalanced datasets lead to difficulties in learning and obtaining accurate predictive models. This paper follows the FAIR paradigm and proposes a technique for the alignment of class distribution, which enables improving image classification performance in imbalanced data and ensuring data reuse. The experiments on the acne disease dataset support that the proposed framework outperforms the baselines and enable to achieve up to 5% improvement in image classification.

The Use of Machine Learning for Accessing Landslide Susceptibility Class: Study Case of Kecamatan Pacet, Kabupaten Mojokerto

Kecamatan Pacet, Kabupaten Mojokerto is one of an area with many landslide events in East Java Province. As a mitigation effort, this research aimed to map the landslide susceptibility class distribution of the research area. This research applied a machine learning analysis technic which combined Frequency Ratio (FR) and Logistic Regression (LR) models to assess the landslide susceptibility class distribution. FR bivariate analysis is used to normalized the data and to identify the influence significancy on each class of triggering factors. LR multivariate analysis is applied to generate the landslide probability (susceptibility) and to show the influence significancy of each triggering factor to landslide events. There are 12 triggering factors to landslide used in this research, which is: TPI, TWI, SPI, slope, aspect, elevation, profile curvature, distance to drainage, geological unit, rainfall, land use, and distance to the road. This research has 383 landslides and 383 non-landslide events as the data sample based on field survey, BPBD Kabupaten Mojokerto, and Google Earth Pro imagery interpretation. The proportion of dataset training and testing is 70% and 30%, which generated from the data inventory. This research used ROC analysis to validate the landslide susceptibility model. The result showed that the landslide susceptibility model has an AUC value of 0.91, which indicated that the model has high accuracy.

SEMI-SUPERVISED SEGMENTATION OF CONCRETE AGGREGATE USING CONSENSUS REGULARISATION AND PRIOR GUIDANCE

In order to leverage and profit from unlabelled data, semi-supervised frameworks for semantic segmentation based on consistency training have been proven to be powerful tools to significantly improve the performance of purely supervised segmentation learning. However, the consensus principle behind consistency training has at least one drawback, which we identify in this paper: imbalanced label distributions within the data. To overcome the limitations of standard consistency training, we propose a novel semi-supervised framework for semantic segmentation, introducing additional losses based on prior knowledge. Specifically, we propose a lightweight architecture consisting of a shared encoder and a main decoder, which is trained in a supervised manner. An auxiliary decoder is added as additional branch in order to make use of unlabelled data based on consensus training, and we add additional constraints derived from prior information on the class distribution and on auto-encoder regularisation. Experiments performed on our concrete aggregate dataset presented in this paper demonstrate the effectiveness of the proposed approach, outperforming the segmentation results achieved by purely supervised segmentation and standard consistency training.

Decision Tree dalam Menganalisis Data Pengunjung Wisata Danau Poso untuk Pengambilan Keputusan

This study uses a classification system in managing its data. In classification there are several methods provided, one of which is the decision tree method with the C4.5 algorithm this method means a decision tree where the structure is the same as a flowchart where each node signifies an attribute test, each branch presents the test results and the leaf node represents the class or class distribution. The data used is the data of Lake Poso Tourism visitors from 2009 to 2020, then the method used in this study is divided into several stages, namely the data being studied, analyzing the data, transforming data and designing a decision tree with the C4.5 algorithm. The results achieved from this study are that the number of visitors more than 28,984 has a description of "Much" which is dominated by local tourists, while the value with the name "Less" is in foreign tourists. This is one of the important points in determining the right strategy for developing tourism in Lake Poso.

Application of natural resources and their economic use in the motivational phase

Explores the forms of class distribution of natural resources in the motivation phase. Analyzes the use of natural resources at this stage in a class-chapter-topic sequence. Investigates the means by which these issues are addressed in the motivation block. Analyzes what knowledge, skills and habits the students have acquired at this stage. It determines the methods and means by which these processes are carried out. Key words: natural resources, news, human economy, analyze, motivation block,text, map-scheme, knowledge, skill, memory, environmental challenges