Few-Shot Learning has emerged as a topic that maximises DNN performance based on very few samples. In Generalised Few-Shot Learning, a model has to learn new few-shot classes while recalling earlier large-scale training classes. Learning the new classes leads to a drop in performance on the base ones. In this work, we identify and explore the parallels between Generalised Few-Shot Object Detection (G-FSOD) and Continual Learning (CL), focusing on two areas in particular: gradient manipulation methods and sampling strategies. Through extensive experimentation we demonstrate that gradient manipulation methods appear to be no better than existing techniques and do not improve performance, but actually harm performance unless the gradients are averaged. Our investigations into sampling strategies consider a number of aspects: the impact of removing the base limit and the effectiveness of different distance measures (with respect to a class prototype) for sample selection. Our experiments into these aspects reveal illuminating insights into their impact on Average Precision on the COCO and VOC datasets. Consequently, we suggest that G-FSOD research focus on the replay aspect and investigate other sampling strategies.

The natural process of understanding speech involves combining auditory and visual cues. CNN based lip reading systems have become very popular in recent years. However, many of these systems consider lipreading to be a black box problem, with limited detailed performance analysis. In this paper, we performed transfer learning by training the Inception v3 CNN model, which has pre-trained weights produced from IMAGENET, with the GRID corpus, delivering good speech recognition results, with 0.61 precision, 0.53 recall, and 0.51 F1-score. The lip reading model was able to automatically learn pertinent features, demonstrated using visualisation, and achieve speaker-independent results comparable to human lip readers on the GRID corpus. We also identify limitations that match those of humans, therefore limiting potential deep learning performance in real world situations.

Deep speaker embedding has demonstrated state-of-the-art performance in speaker recognition tasks. However, one potential issue with this approach is that the speaker vectors derived from deep embedding models tend to be non-Gaussian for each individual speaker, and non-homogeneous for distributions of different speakers. These irregular distributions can seriously impact speaker recognition performance, especially with the popular PLDA scoring method, which assumes homogeneous Gaussian distribution. In this article, we argue that deep speaker vectors require deep normalization, and propose a deep normalization approach based on a novel discriminative normalization flow (DNF) model. We demonstrate the effectiveness of the proposed approach with experiments using the widely used SITW and CNCeleb corpora. In these experiments, the DNF-based normalization delivered substantial performance gains and also showed strong generalization capability in out-of-domain tests.

Deep neural models, particularly the long short-term memory recurrent neural network (LSTM-RNN) model, have shown great potential for language identification (LID). However, the use of phonetic information has been largely overlooked by most existing neural LID methods, although this information has been used very successfully in conventional phonetic LID systems. We present a phonetic temporal neural model for LID, which is an LSTM-RNN LID system that accepts phonetic features produced by a phone-discriminative DNN as the input, rather than raw acoustic features. This new model is similar to traditional phonetic LID methods, but the phonetic knowledge here is much richer: It is at the frame level and involves compacted information of all phones. Our experiments conducted on the Babel database and the AP16-OLR database demonstrate that the temporal phonetic neural approach is very effective, and significantly outperforms existing acoustic neural models. It also outperforms the conventional i-vector approach on short utterances and in noisy conditions.

Jiyuan Zhang, Yang Feng, Dong Wang, Yang Wang, Andrew Abel, Shiyue Zhang, Andi Zhang. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2017.

Neural machine translation (NMT) has achieved notable success in recent times, however it is also widely recognized that this approach has limitations with handling infrequent words and word pairs.This paper presents a novel memoryaugmented NMT (M-NMT) architecture, which stores knowledge about how words (usually infrequently encountered ones) should be translated in a memory and then utilizes them to assist the neural model.We use this memory mechanism to combine the knowledge learned from a conventional statistical machine translation system and the rules learned by an NMT system, and also propose a solution for out-of-vocabulary (OOV) words based on this framework.Our experiments on two Chinese-English translation tasks demonstrated that the M-NMT architecture outperformed the NMT baseline by 9.0 and 2.7 BLEU points on the two tasks, respectively.Additionally, we found this architecture resulted in a much more effective OOV treatment compared to competitive methods.

Since there is good potential for practical applications such as encrypted image authentication, content owner identification and privacy protection, reversible data hiding in encrypted image (RDHEI) has attracted increasing attention in recent years. In this paper, we propose and evaluate a new separable RDHEI framework. Additional data can be embedded into a cipher image previously encrypted using Josephus traversal and a stream cipher. A block histogram shifting (BHS) approach using self-hidden peak pixels is adopted to perform reversible data embedding. Depending on the keys held, legal receivers can extract only the embedded data with the data hiding key, or, they can decrypt an image very similar to the original with the decryption key. They can extract both the embedded data and recover the original image error-free if both keys are available. The results demonstrate that higher embedding payload, better quality of decrypted-marked image and error-free image recovery are achieved.

A key component in the processing of speech is the division of longer input sounds into a number of smaller sections. For speech interpretation it is generally easier to classify single sections. Similarly, when processing speech for other purposes (e.g. speech filtering), it can be easier and more relevant to process individual phonemes. Here, we propose a biologically inspired speech segmentation technique that filters the speech into multiple bandpassed channels using a Gammatone filterbank, and then uses an essentially energy-based spike coding technique in order to find the onsets and offsets present in an audio signal. These onsets and offsets are then processed using leaky integrate-and-fire neurons, and the spikes from these used to determine the speech segmentation. We evaluate this new system using a quantitative evaluation metric, and the promising results of segmentation of both clean speech and speech in noise demonstrate the effectiveness of this technique.