Tarek A. Lahlou

Tarek A. Lahlou, Alan V. Oppenheim

This paper presents two novel regularization methods motivated in part by the geometric significance of biorthogonal bases in signal processing applications. These methods, in particular, draw upon the structural relevance of orthogonality and biorthogonality principles and are presented from the perspectives of signal processing, convex programming, continuation methods and nonlinear projection operators. Each method is specifically endowed with either a homotopy or tuning parameter to facilitate tradeoff analysis between accuracy and numerical stability. An example involving a basis comprised of real exponential signals illustrates the utility of the proposed methods on an ill-conditioned inverse problem and the results are compared to standard regularization techniques from the signal processing literature.

Oluwatobi Olabiyi, Erik T. Mueller, Christopher Larson et al.

Open domain neural dialogue models, despite their successes, are known to produce responses that lack relevance, diversity, and in many cases coherence. These shortcomings stem from the limited ability of common training objectives to directly express these properties as well as their interplay with training datasets and model architectures. Toward addressing these problems, this paper proposes bootstrapping a dialogue response generator with an adversarially trained discriminator. The method involves training a neural generator in both autoregressive and traditional teacher-forcing modes, with the maximum likelihood loss of the auto-regressive outputs weighted by the score from a metric-based discriminator model. The discriminator input is a mixture of ground truth labels, the teacher-forcing outputs of the generator, and distractors sampled from the dataset, thereby allowing for richer feedback on the autoregressive outputs of the generator. To improve the calibration of the discriminator output, we also bootstrap the discriminator with the matching of the intermediate features of the ground truth and the generator's autoregressive output. We explore different sampling and adversarial policy optimization strategies during training in order to understand how to encourage response diversity without sacrificing relevance. Our experiments shows that adversarial bootstrapping is effective at addressing exposure bias, leading to improvement in response relevance and coherence. The improvement is demonstrated with the state-of-the-art results on the Movie and Ubuntu dialogue datasets with respect to human evaluations and BLUE, ROGUE, and distinct n-gram scores.

Chris Larson, Tarek Lahlou, Diana Mingels et al.

Speech processing systems rely on robust feature extraction to handle phonetic and semantic variations found in natural language. While techniques exist for desensitizing features to common noise patterns produced by Speech-to-Text (STT) and Text-to-Speech (TTS) systems, the question remains how to best leverage state-of-the-art language models (which capture rich semantic features, but are trained on only written text) on inputs with ASR errors. In this paper, we present Telephonetic, a data augmentation framework that helps robustify language model features to ASR corrupted inputs. To capture phonetic alterations, we employ a character-level language model trained using probabilistic masking. Phonetic augmentations are generated in two stages: a TTS encoder (Tacotron 2, WaveGlow) and a STT decoder (DeepSpeech). Similarly, semantic perturbations are produced by sampling from nearby words in an embedding space, which is computed using the BERT language model. Words are selected for augmentation according to a hierarchical grammar sampling strategy. Telephonetic is evaluated on the Penn Treebank (PTB) corpus, and demonstrates its effectiveness as a bootstrapping technique for transferring neural language models to the speech domain. Notably, our language model achieves a test perplexity of 37.49 on PTB, which to our knowledge is state-of-the-art among models trained only on PTB.

Tarek A. Lahlou, Thomas A. Baran

This primary purpose of this paper is to succinctly state a number of verifiable and tractable sufficient conditions under which a particular class of conservative signal processing structures may be readily used to solve a companion class of constraint satisfaction problems using both synchronous and asynchronous implementation protocols. In particular, the mentioned class of structures is shown to have desirable convergence and robustness properties with respect to various uncertainties involving communication and processing delays. Essential ingredients to the arguments herein involve blending together functional composition methods, conservation principles, asynchronous signal processing implementation protocols, and methods of homotopy. Numerical experiments complement the theoretical presentation and connections to optimization theory are made.

Tarek A. Lahlou, Anuran Makur

In this paper, we study the problem of transient signal analysis. A signal-dependent algorithm is proposed which sequentially identifies the countable sets of decay rates and expansion coefficients present in a given signal. We qualitatively compare our method to existing techniques such as orthogonal exponential transforms generated from orthogonal polynomial classes. The presented algorithm has immediate utility to signal processing applications wherein the decay rates and expansion coefficients associated with a transient signal convey information. We also provide a functional interpretation of our parameter extraction method via signal approximation using monomials over the unit interval from the perspective of biorthogonal constraint satisfaction.