Warning: statistical benchmarking is addictive. Kicking the habit in machine learning

Algorithm performance evaluation is so entrenched in the machine learning community that one could call it an addiction. Like most addictions, it is harmful and very difficult to give up. It is harmful because it has serious limitations. Yet, we have great faith in practicing it in a ritualistic manner: we follow a fixed set of rules telling us the measure, the data sets and the statistical test to use. When we read a paper, even as reviewers, we are not sufficiently critical of results that follow these rules. Here, we will debate what are the limitations and how to best address them. This article may not cure the addiction but hopefully it will be a good first step along that road.     

Eliminating Synaptic Ribbons from Rods and Cones Halves the Releasable Vesicle Pool and Slows Down Replenishment

Glutamate release from rod and cone photoreceptor cells involves presynaptic ribbons composed largely of the protein RIBEYE. To examine roles of ribbons in rods and cones, we studied mice in which GCamP3 replaced the B-domain of RIBEYE. We discovered that ribbons were absent from rods and cones of both knock-in mice possessing GCamP3 and conditional RIBEYE knockout mice. The mice lacking ribbons showed reduced temporal resolution and contrast sensitivity assessed with optomotor reflexes. ERG recordings showed 50% reduction in scotopic and photopic b-waves. The readily releasable pool (RRP) of vesicles in rods and cones measured using glutamate transporter anion currents (I_A(glu)) was also halved. We also studied the release from cones by stimulating them optogenetically with ChannelRhodopsin2 (ChR2) while recording postsynaptic currents in horizontal cells. Recovery of the release from paired pulse depression was twofold slower in the rods and cones lacking ribbons. The release from rods at −40 mV in darkness involves regularly spaced multivesicular fusion events. While the regular pattern of release remained in the rods lacking ribbons, the number of vesicles comprising each multivesicular event was halved. Our results support conclusions that synaptic ribbons in rods and cones expand the RRP, speed up vesicle replenishment, and augment some forms of multivesicular release. Slower replenishment and a smaller RRP in photoreceptors lacking ribbons may contribute to diminished temporal frequency responses and weaker contrast sensitivity.     

流行快报

2011春夏我们沉浸在寻找怀旧风格和款式的乐趣中,但会用更松弛的感觉改良它们。大胆的印花和鲜艳的颜色在这样一个朝气蓬勃的季节戏份很重。在本季,我们还可以随意地改变比例及混搭面料。     

Failure detection in automotive light assemblies during vibration endurance testing

As the automotive industry becomes increasingly competitive, parts manufacturers are under extreme pressure to improve the quality of their parts, while at the same time reducing costs. The method currently used to detect failures in automotive light assemblies during vibration endurance testing involves manual inspection only after the test is completed. In most cases, this method does not allow the first onset of failure to be determined. Subsequent damage caused by the ongoing test after the initial failure often obliterates the clues to what failed first, thus diminishing the usefulness of the information collected as input to design improvement iterations. An adaptable, reliable, and low-cost real-time monitoring and diagnostic system that would interrupt the testing operation at the first onset of a failure is desired. This paper describes an accelerometer-based and a microphone-based monitoring system for automotive light assembly failure detection during endurance testing. The vibration and acoustic signal analysis methods are also described. Preliminary results from these two systems show a significant difference between healthy and faulty fog light assemblies, which can be detected. This suggests that there is potential to develop these and potentially other measures for use in an adaptable, reliable, and low-cost real-time monitoring and diagnostic system that would interrupt the testing operation at the first onset of a failure. Further testing is planned to determine the sensitivity of the methods used in this study to detect the earliest onset of failure damage during endurance testing of automotive light assemblies.     

用可编程逻辑创建专用微控制器

微控制器是最原始的芯片系统（SoC)器件，在最近几年，FPGA和PLD厂商提出了“可编程芯片系统”（SOPC）的概念，用可编程逻辑来设计专用硬件如微控制器。创建专用微控制器的要素包括全功能的32位RISC微处理器、现成的外设以及可选的专用外设、IP的各种开发工具。这些要素都是现在的，再结合低成本的可编程器件（PLD），设计者就可以为现有的微控制器增加各种功能。本文讨论设计者在PLD设计专用微控制器时的问题，列举的设计实例包括用户定义的硬件加速专用指令和外设。     

Assessing the performance of crack detection tests for solder joints

This paper presents both modelling and experimental test data to characterise the performance of four non-destructive tests. The focus is on determining the presence and rough magnitude of thermal fatigue cracks within the solder joints for a surface mount resistor on a strip of FR4 PCB. The tests all operate by applying mechanical loads to the PCB and monitoring the strain response at the top of the resistor. The modelling results show that of the four tests investigated, three are sensitive to the presence of a crack in the joint and its magnitude. Hence these tests show promise in being able to detect cracking caused by accelerated testing. The experimental data supports these results although more validation is required.     

Perceptually lossless medical image coding

A novel perceptually lossless coder is presented for the compression of medical images. Built on the JPEG 2000 coding framework, the heart of the proposed coder is a visual pruning function, embedded with an advanced human vision model to identify and to remove visually insignificant/irrelevant information. The proposed coder offers the advantages of simplicity and modularity with bit-stream compliance. Current results have shown superior compression ratio gains over that of its information lossless counterparts without any visible distortion. In addition, a case study consisting of 31 medical experts has shown that no perceivable difference of statistical significance exists between the original images and the images compressed by the proposed coder.     

Recapillarity: Electrochemically Controlled Capillary Withdrawal of a Liquid Metal Alloy from Microchannels

This paper describes the mechanistic details of an electrochemical method to control the withdrawal of a liquid metal alloy, eutectic gallium indium (EGaIn), from microfluidic channels. EGaIn is one of several alloys of gallium that are liquid at room temperature and form a thin (nm scale) surface oxide that stabilizes the shape of the metal in microchannels. Applying a reductive potential to the metal removes the oxide in the presence of electrolyte and induces capillary behavior; we call this behavior “recapillarity” because of the importance of electrochemical reduction to the process. Recapillarity can repeatably toggle on and off capillary behavior by applying voltage, which is useful for controlling the withdrawal of metal from microchannels. This paper explores the mechanism of withdrawal and identifies the applied current as the key factor dictating the withdrawal velocity. Experimental observations suggest that this current may be necessary to reduce the oxide on the leading interface of the metal as well as the oxide sandwiched between the wall of the microchannel and the bulk liquid metal. The ability to control the shape and position of a metal using an applied voltage may prove useful for shape reconfigurable electronics, optics, transient circuits, and microfluidic components.