Search

Purpose: We apply federated learning to train an OCT image classifier simulating a realistic scenario with multiple clients and statistical heterogeneous data distribution where data in the clients lack samples of some categories entirely. Methods: We investigate the effectiveness of FedAvg and FedProx to train an OCT image classification model in a decentralized fashion, addressing privacy concerns associated with centralizing data. We partitioned a publicly available OCT dataset across multiple clients under IID and Non-IID settings and conducted local training on the subsets for each client. We evaluated two federated learning methods, FedAvg and FedProx for these settings. Results: Our experiments on the dataset suggest that under IID settings, both methods perform on par with training on a central data pool. However, the performance of both algorithms declines as we increase the statistical heterogeneity across the client data, while FedProx consistently performs better than FedAvg in the increased heterogeneity settings. Conclusion: Despite the effectiveness of federated learning in the utilization of private data across multiple medical institutions, the large number of clients and heterogeneous distribution of labels deteriorate the performance of both algorithms. Notably, FedProx appears to be more robust to the increased heterogeneity.

Resistivity measurements of TiSe$_2$ typically show only a weak change in gradient at the charge density wave transition at $T_{CDW}\approx$ 200~K, but more prominently feature a broad peak at a lower $T_{peak}\sim$ 165~K, which has remained poorly understood despite decades of research on the material. Here we present quantitative simulations of the resistivity using a simplified parametrization of the normal state band structure, based on recent photoemission data. Our simulations reproduce the overall profile of the resistivity of TiSe$_2$, including its prominent peak, without implementing the CDW at all. We find that the peak in resistivity corresponds to a crossover between a low temperature regime with electron-like carriers only, to a regime around room temperature where thermally activated and highly mobile hole-like carriers dominate the conductivity. Even when implementing substantial modifications to model the CDW below the transition temperature, we find that these thermal population effects still dominate the transport properties of TiSe$_2$.

Though conventional public schools enroll the vast majority of difficult-to-educate students, many such students who are not accepted by public schools end up at private schools. This report describes the role of private institutions in educating these students, incorporating case studies and analyzing various institutional arrangements. Part 1 introduces the subject; describes private programs (special education programs for students with disabilities, education for at-risk students, and education for incarcerated youth); and discusses alternative arrangements (charter schools, full inclusion, and cooperative agreements). Part 2 focuses on program analysis (performance measures, financial accountability, and student access) and policy issues (cost drivers, implications for school choice, and private schools, special education, and the courts). Part 3 looks at case studies of six Michigan facilities to indicate the diversity and capability of Michigan's nongovernment providers of education for difficult-to-educate students. Four appendixes present data on: financial incentives; states with legislation allowing public schools to contract for alternative education for at-risk students; defining characteristics of educational placements; and service providers and key contacts. (Contains 144 endnotes.) (SM)

In December 1994, San Diego City Schools' (California) Board of Education voted to discontinue competitive contracting for busing and expand its in-house transportation department, believing the new arrangement would save money. This report argues that the decision was based on a flawed financial analysis prepared by the district's in-house transportation department. Instead of looking at total cost, the analysis considered only the marginal cost of expanding in-house services. The report argues that in-house costs for school-bus service were 80 percent higher than private-contract providers. The district incurred higher costs in the areas of labor costs, utilization of resources, and fleet acquisition and owning costs. Private contractors operated more efficiently than the Transportation Services Department in three areas: lower labor costs, better utilization of vehicles over useful life, and lower vehicle purchase price. By eliminating contract service, the school board gave the Transportation Services Department (TSD) little incentive to control costs. The report recommends that the San Diego Board of Education: (1) halt further expansion plans until costs are fully identified; (2) invite an independent auditor to evaluate total costs of the TSD and the cost of service alternatives; and (3) implement fully allocated costing methods into the district's accounting system. Six tables and two figures are included. Appendices contain cost-analysis strategies, definitions, a list of recommended reading, tips for creating a competitive environment for transportation services, and discussion of the continuous-improvement strategy. (LMI)

In response to declining student performance in the Milwaukee Public Schools (MPS), the Wisconsin State Legislature and the private sector each created programs to give school choice to low-income students. In 1990-91 the Legislature implemented the Milwaukee Parental Choice Program (MPCP). About 750 students received government-funded tuition vouchers in 1993-94 to attend any one of a dozen nonreligious private schools in Milwaukee. In 1992, business and religious organizations joined to establish Partners Advancing Values in Education (PAVE), a privately funded school-choice program for low-income students. Unlike the MPCP, PAVE's tuition scholarships may be used at any private school in Milwaukee, including religious schools. PAVE served approximately 2,370 students enrolled in 102 private schools during the 1993-94 school year. This document presents findings of a study that compared the two programs. Data were obtained from parent surveys and student-academic records. Findings indicate that PAVE students outperformed both MPCP and MPS students on standardized tests, and that 57 percent of PAVE families were headed by a single parent. PAVE students who had previously been enrolled in private schools and PAVE students who had previously been enrolled in public schools held nearly identical demographic characteristics. However, PAVE students from private schools performed significantly better on standardized tests. Parents from both programs cited educational quality as the most important reason for choosing a school, followed by discipline and general atmosphere. PAVE parents and MPCP parents were both very satisfied with the amount their children learned in school, though PAVE parents were slightly more satisfied. Students of diverse religions used PAVE scholarships, but PAVE tended to enroll more white students. PAVE schools also had better retention rates than MPCP schools. It is recommended that promoters of school-choice programs improve information dissemination to reach the "mainstream" low-income public-school parents. Fifteen tables are included. Appendices contain survey data. (LMI)

This document examines the implications and feasibility of school-voucher programs for the California education system. Following the executive summary and introduction, section 2 describes privately provided voucher programs, specifically: the Educational CHOICE Charitable Trust (Indianapolis, Indiana); The Children's Educational Opportunity (CEO) Foundation (San Antonio, Texas); Children's Education Foundation (Atlanta, Georgia); and Partners Advancing Values in Education (PAVE) (Milwaukee, Wisconsin). The third section presents findings of the Milwaukee Parental Choice Program (MPCP), a government-supported voucher program. Findings indicate that more low-income parents have applied to the PAVE program than to the MPCP because of fewer restrictions on access. Lessons learned about school choice are that: (1) regulations block access; (2) parents in all five programs identified educational quality as the most important consideration in choosing a school; and (3) bias can operate against minority students. Section 5 describes methodologies used by states with privately funded voucher programs. Recommendations for establishing a private-voucher program in California are offered in the sixth section. The conclusion recommends that state governments grant tax credits to organizations providing tuition vouchers and reduce business regulations affecting private schools. Six tables, a contact list, and list of target areas, are included. (LMI)

The private-sector provision of school infrastructure in the form of satellite schools is discussed in this paper. Following the introduction, section 2 presents a case study of Satellite Learning Centers in Dade County, Florida, in which the schools operate as public schools on business worksites. The host-business, American Bankers Insurance Group (ABIG) contributes land, building space, and some operating expenses; the school district supplies the rest--teachers, supplies, curriculum, and administration. Benefits include the savings of millions of dollars in public school infrastructure and transportation costs, increased student performance and attendance, improved student/parent/teacher interaction, decreased absenteeism among working parents, and greater teacher career advancement opportunities. Section 3 describes the Hidden Valley Satellite School, a partnership between Hewlett-Packard Co. and the Santa Rosa City School District. The fourth section outlines potential benefits of satellite schools, which offer a solution to school overcrowding at a minimal cost relative to other options. An obstacle is the Field Act, California's seismic-safety standards code that prevents satellite schools from occupying existing office buildings. Recommendations are made for school districts, business partners, and the California State Government. Two tables are included. Appendices contain the ABIG-Dade County contract, various projections and program descriptions, and the Hidden Valley contract. (Contains 46 references.) (LMI)

Findings of a study that investigated the feasibility of educational vouchers for California education and their budgetary impact are presented in this paper. In a survey of 300 parents in the Los Angeles Unifies School District with children under the age of 16 who do or will attend public schools, 52 percent said that they would use a $2,600 voucher to send their children to private school. Support was highest among minorities, particularly African-Americans. Budgetary implications are that a 52 percent use of vouchers statewide could reduce California's education budget by $3 billion, which could be reinvested in public schools or for other general foundation purposes. However, cost savings at the state level may not map the underlying cost structure of individual school districts. Because public school operating costs are composed of both fixed and variable costs, the number of students who leave the public school must be great enough to reduce fixed costs before the voucher program becomes cost-effective for individual school districts. Two tables and one graph are included. (LMI)