arXiv Science⌕ Search

arXiv subjects

Manjot Kaur

Publications and source records attributed to Manjot Kaur.

3 recordsLinked to original sources

Sensitivity-Gated Accumulator Caching for On-Device Learning of Low-Bit Hand Networks

Small hand-tracking networks are deployed as 4- and 8-bit codes, and they should keep adapting to a user's camera after deployment, on the device and within roughly the memory of the model itself. Quantization-aware training cannot do this cheaply, because it keeps a 32-bit master copy of every weight together with its optimizer state, about 96 bits per weight. Training the codes directly with stochastic rounding needs no copy but injects rounding noise at every step, and the methods that remove this noise keep a residual buffer for every weight. We introduce sensitivity-gated accumulator caching (SGAC), which keeps residual state for only a small and changing set of weight groups. A set-associative cache of 8-bit accumulators collects the sub-step updates of the groups with the largest Fisher-weighted squared code step and writes them into the codes as whole steps; an evicted line is flushed by stochastic rounding, and a group without a line does not move until it is admitted. We prove that the write-back is unbiased and show, under a local quadratic model, that rounding a sub-step update costs loss of the same order as the update's own gain, so that rounding pays only where the expected gradient is large relative to the curvature. On a 383k-parameter hand-keypoint network, SGAC trains with 8.1 bits of state per weight. Scored on the 4/8-bit network that the device runs after adaptation, it has the highest mean accuracy of nine deployed low-bit networks on two simulated cameras. On twelve ImageNet-C corruptions it has the highest average of corrupted and clean accuracy of seven deployed methods, ahead of the PEQA-style, ECO-style and direct low-bit baselines with confidence intervals that exclude zero, and it matches an 8-bit accumulator kept for every weight with 56% of that state.

cs.CV↗

Low temperature carrier transport mechanism and photoconductivity of WSe2

This work reports the electrical transport and temperature-dependent photoconductivity in tungsten diselenide (WSe2) thin films. The electrical conductivity analysis shows the presence of the three regions with temperature variation. At lower temperatures (<190K), carriers become localized to small regions in the film due to the Mott hopping mechanism. The middle region (190 to 273 K) follows Seto parameters and obtained low barrier height (0.0873 eV) may be responsible for the improved carrier mobility. At higher temperature (>273K) region, thermally activated conduction is dominated with two activation energies of ~138 meV and 98 meV. The peaks obtained in photoluminescent analysis attributes to the presence of mid-bandgap states or defect states which play an important role in the photoconductivity of WSe2. The transient photoconductivity measurements show consistent temperature-dependent behaviour. The effect of light intensity and wavelength variation on the photoconductivity of WSe2 thin films is also discussed. The photocurrent is 1.19*10-5 A at 125 K while at 350 K it was observed to be 3.12*10-4 A. The light-on/off current cycles show that the current can recover to its initial state which points to the stable and outstanding reversible properties of the WSe2 thin film device to be used in photodetector applications.

cond-mat.mtrl-sci↗

Specially designed B4C/SnO2 nanocomposite for photocatalysis: traditional ceramic with unique properties

Boron carbide: A traditional ceramic material shows unique properties when explored in nano-range. Specially designed boron based nanocomposite has been synthesized by reflux method. The addition of SnO2 in base matrix increase the defect states in boron carbide and shows unique catalytic properties. The calculated texture coefficient and Nelson Riley factor shows that the synthesized nanocomposite have very high defect states. Also this composite is explored for the first time for catalysis degradation of industrial used dyes. The industrial pollutants such as Novacron red and methylene blue dye degradation analysis reveal that the composite is an efficient catalyst. Degradation study shows that 1 g/L catalyst concentration of B4C/SnO2 degrade Novacron red Huntsman dye upto 97.38% approximately in 20 minutes under sunlight irradiation time. This water insoluble catalyst can be recovered and reused.

physics.app-ph↗