Quantifying the computational capability of a nanomagnetic reservoir computing platform with emergent magnetisation dynamics.

Devices based on arrays of interconnected magnetic nano-rings with emergent magnetization dynamics have recently been proposed for use in reservoir computing applications, but for them to be computationally useful it must be possible to optimise their dynamical responses. Here, we use a phenomenological model to demonstrate that such reservoirs can be optimised for classification tasks by tuning hyperparameters that control the scaling and input-rate of data into the system using rotating magnetic fields. We use task-independent metrics to assess the rings' computational capabilities at each set of these hyperparameters and show how these metrics correlate directly to performance in spoken and written digit recognition tasks. We then show that these metrics, and performance in tasks, can be further improved by expanding the reservoir's output to include multiple, concurrent measures of the ring arrays' magnetic states.

The role of low Gd concentrations on magnetisation behaviour in rare earth:transition metal alloy films.

The magnetisation reversal behaviour as a function of composition was studied in low rare earth concentration alloys. 30 nm thick rare earth:transition-metal films of composition GdxCo100-x, GdxFe100-x and Gdx(Co50Fe50)100-x were prepared by magnetron sputtering, where x ranged from 4 to 13 atomic%. Magnetisation behaviour was studied using MOKE and Hall hysteresis measurements. The magnetic reversal behaviour as a function of Gd content is strongly dependent on the transition metal. With increasing Gd content the film structure transitions from crystalline to amorphous and the saturation magnetisation decreases linearly. For GdCo, the reversal field, Hc, increases by less than a factor of two with Gd doping of 11%, while for Fe, the coercivity falls by a factor of ten with 8% Gd. This may be attributed to changes in the crystalline morphology. GdCoFe shows a similar trend with Gd doping for the in-plane reversal field to that of GdFe. With 13% Gd in Fe there is evidence indicating the presence of a weak perpendicular magnetic anisotropy, PMA. With Gd doping the anomalous Hall resistivity of Co, Fe and CoFe increases significantly with the largest increase observed for GdCoFe.

Chemical characterisation of Hoodia gordonii extract.

The chemical composition of a solvent extract of Hoodia gordonii termed 'H.gordonii extract' has been characterised by hyphenated chromatographic methods and traditional analytical techniques. The extract consists of a mixture of steroid glycosides, fatty acids, plant sterols and polar organic material. High performance liquid chromatography (HPLC) with ultra violet (UV) and mass spectrometric (MS) detection was used to quantify and confirm the identity of a number of steroid glycosides (73.7% w/w) present in the extract. Gas chromatography (GC) with MS and flame ionisation detection (FID) was applied to determine the fatty acid (3.12% w/w) sterol (0.39% w/w) and alcohol (0.03% w/w) content of a saponified sample of the extract. Polar organic material was quantified by gravimetric methodology using C(18) SPE separation and was determined to be a minimum of 3% w/w. Moisture content was measured by Karl Fischer coulometric titration (0.81% w/w). The protein content was investigated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) with SYPRO Ruby staining and a negative result was determined with a limit of detection of <0.001%w/w of protein per band. The chemical composition of the extract remained stable for 19 months when stored in re-sealable plastic bags at ambient (21-24°C) temperature and <60% relative humidity.

Designing for Feel: Contrasts between Human and Automated Parametric Capture of Knob Physics.

We examine a crucial aspect of a tool intended to support designing for feel: the ability of an objective physical-model identification method to capture perceptually relevant parameters, relative to human identification performance. The feel of manual controls, such as knobs, sliders, and buttons, becomes critical when these controls are used in certain settings. Appropriate feel enables designers to create consistent control behaviors that lead to improved usability and safety. For example, a heavy knob with stiff detents for a power plant boiler setting may afford better feedback and safer operations, whereas subtle detents in an automobile radio volume knob may afford improved ergonomics and driver attention to the road. To assess the quality of our identification method, we compared previously reported automated model captures for five real mechanical reference knobs with captures by novice and expert human participants who were asked to adjust four parameters of a rendered knob model to match the feel of each reference knob. Participants indicated their satisfaction with the matches their renderings produced. We observed similar relative inertia, friction, detent strength, and detent spacing parameterizations by human experts and our automatic estimation methods. Qualitative results provided insight on users' strategies and confidence. While experts (but not novices) were better able to ascertain an underlying model in the presence of unmodeled dynamics, the objective algorithm outperformed all humans when an appropriate physical model was used. Our studies demonstrate that automated model identification can capture knob dynamics as perceived by a human, and they also establish limits to that ability; they comprise a step towards pragmatic design guidelines for embedded physical interfaces in which methodological expedience is informed by human perceptual requirements.