Controlled experiments on the web: survey and practical guide

The web provides an unprecedented opportunity to evaluate ideas quickly using controlled experiments, also called randomized experiments, A/B tests (and their generalizations), split tests, Control/Treatment tests, MultiVariable Tests (MVT) and parallel flights. Controlled experiments embody the best scientific design for establishing a causal relationship between changes and their influence on user-observable behavior. We provide a practical guide to conducting online experiments, where end-users can help guide the development of features. Our experience indicates that significant learning and return-on-investment (ROI) are seen when development teams listen to their customers, not to the Highest Paid Person’s Opinion (HiPPO). We provide several examples of controlled experiments with surprising results. We review the important ingredients of running controlled experiments, and discuss their limitations (both technical and organizational). We focus on several areas that are critical to experimentation, including statistical power, sample size, and techniques for variance reduction. We describe common architectures for experimentation systems and analyze their advantages and disadvantages. We evaluate randomization and hashing techniques, which we show are not as simple in practice as is often assumed. Controlled experiments typically generate large amounts of data, which can be analyzed using data mining techniques to gain deeper understanding of the factors influencing the outcome of interest, leading to new hypotheses and creating a virtuous cycle of improvements. Organizations that embrace controlled experiments with clear evaluation criteria can evolve their systems with automated optimizations and real-time analyses. Based on our extensive practical experience with multiple systems and organizations, we share key lessons that will help practitioners in running trustworthy controlled experiments.     

Mechanism of adsorption of human albumin to titanium in vitro

1. This study explored the hypothesis that atrial natriuretic peptide (ANP), brain natriuretic peptide (BNP), and C-natriuretic peptide (CNP) have differing antiproliferative and antihypertrophic effects on pulmonary artery (PA) and thoracic aorta (TA) smooth-muscle cells (SMCs). 2. Cultured cells were exposed to 5% fetal calf serum (FCS) and angiotensin II (A-II) to induce DNA and protein synthesis, respectively. 3. ANP (10(-7) M) significantly reduced thymidine uptake in TA by 31% +/- 2% (P < or = 0.01) but not in PA (P > or = 0.05). 4. In parallel experiments, BNP (10(-7) M) significantly reduced thymidine uptake in TA (-22% +/- 5%, P < or = 0.01), but not in PA cells (P > or = 0.05). 5. CNP (10(-7) M) did not significantly alter thymidine uptake in TA cells exposed to FCS, but it did significantly reduce uptake in PA (-28.5% +/- 4%) 2(P < or = 0.05). 6. Blunting by ANP (10(-7) M) of the A-II (10(-8) M)-induced increase in protein synthesis was significantly greater in PA than in TA cells. 7. However, BNP and CNP (10(-7) M) exerted similar antihypertrophic effects on TA and PA cells exposed to A-II. 8. The antiproliferative effects of BNP and ANP exceed those of CNP in TA SMCs, but CNP appears to be the most effective antiproliferative agent in PA SMCs. In addition, PA-derived SMCs are more sensitive to the antihypertrophic effects of ANP than TA-derived cells, suggesting phenotypic differences. The findings indicate that the natriuretic peptides may play complementary roles in modulating SMC proliferation and protein synthesis.     

Lessons and Challenges from Mining Retail E-Commerce Data

The architecture of Blue Martini Software's e-commerce suite has supported data collection, data transformation, and data mining since its inception. With clickstreams being collected at the application-server layer, high-level events being logged, and data automatically transformed into a data warehouse using meta-data, common problems plaguing data mining using weblogs (e.g., sessionization and conflating multi-sourced data) were obviated, thus allowing us to concentrate on actual data mining goals. The paper briefly reviews the architecture and discusses many lessons learned over the last four years and the challenges that still need to be addressed. The lessons and challenges are presented across two dimensions: business-level vs. technical, and throughout the data mining lifecycle stages of data collection, data warehouse construction, business intelligence, and deployment. The lessons and challenges are also widely applicable to data mining domains outside retail e-commerce.     

The minimum test set problem for circuits with nonreconvergent fanout

The problem of computing the minimum size of a test set for a combinational circuit is considered. It is known that the minimum test set size of a combinational circuit can be determined in polynomial time for fanout free circuits, while even for circuits with non-reconvergent fanout, the minimum test set size problem is NP-Hard. We extend the class of circuits for which a minimum test set can be constructed in polynomial time to include a class of circuits with fanout, called restricted fanout circuits. Restricted fanout circuits are characterized using an undirected graph describing the structure of the circuit. The graph for these circuits must be free of (undirected) cycles. In addition, the paper demonstrates a novel application of dynamic programming to test generation problems.Formerly with the Electrical Eng. and Computer Science Depts. at the Technion, Haifa 32000, Israel.     

A combined impression technique for a partial implant-supported fixed-detachable restoration

INTRODUCTION: The neuroectodermal tumor of bone constitutes a recently isolated neoplasm, which morphologically resembles the peripheral neuroepithelioma of soft tissue. The diagnosis is made by excluding readily classified small round cell neoplasms and then demonstrating the neural origin of the tumor by means of ultrastructural or immunocytochemical studies. CLINICAL CASE: We report a case of a primary neuroectodermal tumor (PNET) with atypical findings (primary cervical location, massive vertebral body infiltration and isolated neurological symptomatology). Total tumor removal was achieved by means of a one-stage three-level laminectomy (partial C5, total C6-C7), C6 somatectomy, and combined anterior and posterior cervical spinal instrumentation and fusion. CONCLUSIONS: Our case presents three peculiar features: 1) isolated myelopathic symptomatology (first case described); 2) primary isolated massive vertebral cervical involvement with intraspinal epidural extension (first case described); 3) one-stage tumor resection, fusion and instrumentation. Since radiation therapy causes tumor shrinkage but is not curative, radical removal is mandatory even with complex instrumentation procedures.     

alpha-Amylase and salivary albumin adsorption onto titanium, enamel and dentin: an in vivo study

In vivo adsorption of salivary albumin and alpha-amylase onto titanium, enamel and dentin was analysed following their exposure to the oral cavity for 2h. Oral appliances in six adults served as a platform for carrying 4-mm discs of the three materials. Adherent proteins were eluted from the discs and the amounts of salivary albumin and alpha-amylase were measured by an enzyme-linked immunosorbent assay. While significant difference between the adsorption of albumin and alpha-amylase onto enamel as compared with dentin was observed, adsorption onto titanium was significantly lower. A sample of whole saliva was also collected from each participant. The mean total amounts of albumin and alpha-amylase in the participants' whole saliva were 0.03 and 0.54 mg ml-1, respectively. Titanium adsorbed significantly less (4.43%) of the total albumin than did enamel (14.30%) or dentin (18.80%). No significant difference was found in the relative amounts of alpha-amylase adsorbed by the three materials. This significantly selective adsorption of proteins may enable the attachment of specific bacteria and thus alter the composition of the dental plaque and its potential pathogenicity.     

Complementary negabinary arithmetic unit

Algorithms for the basic arithmetic operations in negabinary (base-2) representation are described. The complementary negabinary full adder is defined, and its advantages in the realization of negabinary hardware are pointed out.     

An optimal algorithm for cycle breaking in directed graphs

This paper describes an exact algorithm for the identification of a minimal feedback vertex set in digital circuits. The proposed algorithm makes use of graph reduction and efficient graph partitioning methods based on local properties of digital circuits. It has been implemented and applied to ISCAS-89 benchmark circuits. Previously, non-optimum solutions were found. In other cases, the optimality of the solution could not be established for all circuits. By using the proposed algorithm we obtained the optimum results for all the circuits in a relatively short CPU time.Supported in part by the Technion fund for the promotion of research.