|Two training algorithms for self evolving neural networks are discussed for rule based data analysis. E cient classi cation is achieved with less number of automatically added clusters and application data is analysed by interpreting the trained neural network as a fuzzy rule based system. Learning Vector Quantisation algorithm has been modied acquiring the self evolvement character in the pro...

Vector quantisation is a widely used technique in low-bit rate coding of speech signals, but is highly sensitive to errors in the transmitted codeword, due to noise in the transmission channel. This can degrade speech quality quite dramatically if the reference vector retrieved by the corrupted index is very di erent from the reference vector corresponding to the intended index. The index assig...

To enhance the traditional vector quantisation (VQ) system by adding the watermarking ability, a digital image watermarking scheme, which modifies the VQ indices to carry watermark bits, is presented. This scheme partitions the main codebook into two subcodebooks by referring to the user-key. Then, for each input vector of the cover image, a subcodebook is selected according to the watermark bi...

Colour quantisation algorithms are essential for displaying true colour images using a limited palette of distinct colours. The choice of a good colour palette is crucial as it directly determines the quality of the resulting image. Colour quantisation can also be seen as a clustering problem where the task is to identify those clusters that best represent the colours in an image. In this paper...

In this contribution the impact of finite signal wordlengths on the performance of digital systems for arbitrary sampling rate conversion (ASRC), where input and output sampling rates are derived from independent clock generators, is investigated. For two different approaches to ASRC the noise power due to both, input/output quantisation and multiplication roundoff errors, is determined as a fu...

This paper compares the latency, accuracy, training time and hardware costs of neural networks compressed with our new multi-objective evolutionary algorithm called NEMOKD, quantisation. We evaluate NEMOKD on Intel’s Movidius Myriad X VPU processor, quantisation Xilinx’s programmable Z7020 FPGA hardware. Evolving models increases inference accuracy by up to 82% at cost 38% increased throughput ...

We study the quantisation of complex, finite–dimensional, compact, classical phase spaces C, by explicitly constructing Hilbert–space vector bundles over C. We find that these vector bundles split as the direct sum of two holomorphic vector bundles: the holomorphic tangent bundle T (C), plus a complex line bundle N(C). Quantum states (except the vacuum) appear as tangent vectors to C. The vacuu...

The discrete cosine transform (DCT) is widely applied in modern codecs to remove spatial redundancies, with the resulting DCT coefficients being quantised to achieve compression as well as bit-rate control. In distributed video coding (DVC) architectures like DISCOVER, DCT coefficient quantisation is traditionally performed using predetermined quantisation matrices (QM), which means the compres...

We develop a categorical compositional distributional semantics for Lambek Calculus with Relevant Modality, which has limited version of the contraction and permutation rules. The part is monoidal biclosed category coalgebra modality as defined on Differential Categories. instantiate this to finite dimensional vector spaces linear maps via quantisation functors work three concrete interpretatio...