Wonyong Sung

Minsoo Kim, Sihwa Lee, Janghwan Lee et al.

Generative Language Models (GLMs) have shown impressive performance in tasks such as text generation, understanding, and reasoning. However, the large model size poses challenges for practical deployment. To solve this problem, Quantization-Aware Training (QAT) has become increasingly popular. However, current QAT methods for generative models have resulted in a noticeable loss of accuracy. To counteract this issue, we propose a novel knowledge distillation method specifically designed for GLMs. Our method, called token-scaled logit distillation, prevents overfitting and provides superior learning from the teacher model and ground truth. This research marks the first evaluation of ternary weight quantization-aware training of large-scale GLMs with less than 1.0 degradation in perplexity and achieves enhanced accuracy in tasks like common-sense QA and arithmetic reasoning as well as natural language understanding. Our code is available at https://github.com/aiha-lab/TSLD.

Minsoo Kim, Kyuhong Shim, Seongmin Park et al.

Pre-trained Transformer models such as BERT have shown great success in a wide range of applications, but at the cost of substantial increases in model complexity. Quantization-aware training (QAT) is a promising method to lower the implementation cost and energy consumption. However, aggressive quantization below 2-bit causes considerable accuracy degradation due to unstable convergence, especially when the downstream dataset is not abundant. This work proposes a proactive knowledge distillation method called Teacher Intervention (TI) for fast converging QAT of ultra-low precision pre-trained Transformers. TI intervenes layer-wise signal propagation with the intact signal from the teacher to remove the interference of propagated quantization errors, smoothing loss surface of QAT and expediting the convergence. Furthermore, we propose a gradual intervention mechanism to stabilize the recovery of subsections of Transformer layers from quantization. The proposed schemes enable fast convergence of QAT and improve the model accuracy regardless of the diverse characteristics of downstream fine-tuning tasks. We demonstrate that TI consistently achieves superior accuracy with significantly lower fine-tuning iterations on well-known Transformers of natural language processing as well as computer vision compared to the state-of-the-art QAT methods.

Kyuhong Shim, Jungwook Choi, Wonyong Sung

Transformer-based deep neural networks have achieved great success in various sequence applications due to their powerful ability to model long-range dependency. The key module of Transformer is self-attention (SA) which extracts features from the entire sequence regardless of the distance between positions. Although SA helps Transformer performs particularly well on long-range tasks, SA requires quadratic computation and memory complexity with the input sequence length. Recently, attention map reuse, which groups multiple SA layers to share one attention map, has been proposed and achieved significant speedup for speech recognition models. In this paper, we provide a comprehensive study on attention map reuse focusing on its ability to accelerate inference. We compare the method with other SA compression techniques and conduct a breakdown analysis of its advantages for a long sequence. We demonstrate the effectiveness of attention map reuse by measuring the latency on both CPU and GPU platforms.

Janghwan Lee, Minsoo Kim, Seungcheol Baek et al.

Large Language Models (LLMs) are proficient in natural language processing tasks, but their deployment is often restricted by extensive parameter sizes and computational demands. This paper focuses on post-training quantization (PTQ) in LLMs, specifically 4-bit weight and 8-bit activation (W4A8) quantization, to enhance computational efficiency -- a topic less explored compared to weight-only quantization. We present two innovative techniques: activation-quantization-aware scaling (AQAS) and sequence-length-aware calibration (SLAC) to enhance PTQ by considering the combined effects on weights and activations and aligning calibration sequence lengths to target tasks. Moreover, we introduce dINT, a hybrid data format combining integer and denormal representations, to address the underflow issue in W4A8 quantization, where small values are rounded to zero. Through rigorous evaluations of LLMs, including OPT and LLaMA, we demonstrate that our techniques significantly boost task accuracies to levels comparable with full-precision models. By developing arithmetic units compatible with dINT, we further confirm that our methods yield a 2$\times$ hardware efficiency improvement compared to 8-bit integer MAC unit.

Kyuhong Shim, Wonyong Sung

Transformer-based speech recognition models have achieved great success due to the self-attention (SA) mechanism that utilizes every frame in the feature extraction process. Especially, SA heads in lower layers capture various phonetic characteristics by the query-key dot product, which is designed to compute the pairwise relationship between frames. In this paper, we propose a variant of SA to extract more representative phonetic features. The proposed phonetic self-attention (phSA) is composed of two different types of phonetic attention; one is similarity-based and the other is content-based. In short, similarity-based attention captures the correlation between frames while content-based attention only considers each frame without being affected by other frames. We identify which parts of the original dot product equation are related to two different attention patterns and improve each part with simple modifications. Our experiments on phoneme classification and speech recognition show that replacing SA with phSA for lower layers improves the recognition performance without increasing the latency and the parameter size.

Kyuhong Shim, Wonyong Sung

Phoneme recognition is a very important part of speech recognition that requires the ability to extract phonetic features from multiple frames. In this paper, we compare and analyze CNN, RNN, Transformer, and Conformer models using phoneme recognition. For CNN, the ContextNet model is used for the experiments. First, we compare the accuracy of various architectures under different constraints, such as the receptive field length, parameter size, and layer depth. Second, we interpret the performance difference of these models, especially when the observable sequence length varies. Our analyses show that Transformer and Conformer models benefit from the long-range accessibility of self-attention through input frames.

Chanwoo Kim, Sathish Indurti, Jinhwan Park et al.

This paper proposes a new regularization algorithm referred to as macro-block dropout. The overfitting issue has been a difficult problem in training large neural network models. The dropout technique has proven to be simple yet very effective for regularization by preventing complex co-adaptations during training. In our work, we define a macro-block that contains a large number of units from the input to a Recurrent Neural Network (RNN). Rather than applying dropout to each unit, we apply random dropout to each macro-block. This algorithm has the effect of applying different drop out rates for each layer even if we keep a constant average dropout rate, which has better regularization effects. In our experiments using Recurrent Neural Network-Transducer (RNN-T), this algorithm shows relatively 4.30 % and 6.13 % Word Error Rates (WERs) improvement over the conventional dropout on LibriSpeech test-clean and test-other. With an Attention-based Encoder-Decoder (AED) model, this algorithm shows relatively 4.36 % and 5.85 % WERs improvement over the conventional dropout on the same test sets.

Iksoo Choi, Wonyong Sung

As sleep disorders are becoming more prevalent there is an urgent need to classify sleep stages in a less disturbing way.In particular, sleep-stage classification using simple sensors, such as single-channel electroencephalography (EEG), electrooculography (EOG), electromyography (EMG), or electrocardiography (ECG) has gained substantial interest. In this study, we proposed a sleep model that predicts the next sleep stage and used it to improve sleep classification accuracy. The sleep models were built using sleep-sequence data and employed either statistical $n$-gram or deep neural network-based models. We developed beam-search decoding to combine the information from the sensor and the sleep models. Furthermore, we evaluated the performance of the $n$-gram and long short-term memory (LSTM) recurrent neural network (RNN)-based sleep models and demonstrated the improvement of sleep-stage classification using an EOG sensor. The developed sleep models significantly improved the accuracy of sleep-stage classification, particularly in the absence of an EEG sensor.

Kyuhong Shim, Hyewon Bae, Wonyong Sung

The performance of automatic speech recognition (ASR) models can be greatly improved by proper beam-search decoding with external language model (LM). There has been an increasing interest in Korean speech recognition, but not many studies have been focused on the decoding procedure. In this paper, we propose a Korean tokenization method for neural network-based LM used for Korean ASR. Although the common approach is to use the same tokenization method for external LM as the ASR model, we show that it may not be the best choice for Korean. We propose a new tokenization method that inserts a special token, SkipTC, when there is no trailing consonant in a Korean syllable. By utilizing the proposed SkipTC token, the input sequence for LM becomes very regularly patterned so that the LM can better learn the linguistic characteristics. Our experiments show that the proposed approach achieves a lower word error rate compared to the same LM model without SkipTC. In addition, we are the first to report the ASR performance for the recently introduced large-scale 7,600h Korean speech dataset.

Kyuhong Shim, Iksoo Choi, Wonyong Sung et al.

While Transformer-based models have shown impressive language modeling performance, the large computation cost is often prohibitive for practical use. Attention head pruning, which removes unnecessary attention heads in the multihead attention, is a promising technique to solve this problem. However, it does not evenly reduce the overall load because the heavy feedforward module is not affected by head pruning. In this paper, we apply layer-wise attention head pruning on All-attention Transformer so that the entire computation and the number of parameters can be reduced proportionally to the number of pruned heads. While the architecture has the potential to fully utilize head pruning, we propose three training methods that are especially helpful to minimize performance degradation and stabilize the pruning process. Our pruned model shows consistently lower perplexity within a comparable parameter size than Transformer-XL on WikiText-103 language modeling benchmark.

Yoonho Boo, Sungho Shin, Jungwook Choi et al.

The quantization of deep neural networks (QDNNs) has been actively studied for deployment in edge devices. Recent studies employ the knowledge distillation (KD) method to improve the performance of quantized networks. In this study, we propose stochastic precision ensemble training for QDNNs (SPEQ). SPEQ is a knowledge distillation training scheme; however, the teacher is formed by sharing the model parameters of the student network. We obtain the soft labels of the teacher by changing the bit precision of the activation stochastically at each layer of the forward-pass computation. The student model is trained with these soft labels to reduce the activation quantization noise. The cosine similarity loss is employed, instead of the KL-divergence, for KD training. As the teacher model changes continuously by random bit-precision assignment, it exploits the effect of stochastic ensemble KD. SPEQ outperforms the existing quantization training methods in various tasks, such as image classification, question-answering, and transfer learning without the need for cumbersome teacher networks.

Wonyong Sung, Iksoo Choi, Jinhwan Park et al.

The stochastic gradient descent (SGD) method is most widely used for deep neural network (DNN) training. However, the method does not always converge to a flat minimum of the loss surface that can demonstrate high generalization capability. Weight noise injection has been extensively studied for finding flat minima using the SGD method. We devise a new weight-noise injection-based SGD method that adds symmetrical noises to the DNN weights. The training with symmetrical noise evaluates the loss surface at two adjacent points, by which convergence to sharp minima can be avoided. Fixed-magnitude symmetric noises are added to minimize training instability. The proposed method is compared with the conventional SGD method and previous weight-noise injection algorithms using convolutional neural networks for image classification. Particularly, performance improvements in large batch training are demonstrated. This method shows superior performance compared with conventional SGD and weight-noise injection methods regardless of the batch-size and learning rate scheduling algorithms.

Yoonho Boo, Sungho Shin, Wonyong Sung

Quantized deep neural networks (QDNNs) are necessary for low-power, high throughput, and embedded applications. Previous studies mostly focused on developing optimization methods for the quantization of given models. However, quantization sensitivity depends on the model architecture. Therefore, the model selection needs to be a part of the QDNN design process. Also, the characteristics of weight and activation quantization are quite different. This study proposes a holistic approach for the optimization of QDNNs, which contains QDNN training methods as well as quantization-friendly architecture design. Synthesized data is used to visualize the effects of weight and activation quantization. The results indicate that deeper models are more prone to activation quantization, while wider models improve the resiliency to both weight and activation quantization. This study can provide insight into better optimization of QDNNs.

Sungho Shin, Yoonho Boo, Wonyong Sung

Designing a deep neural network (DNN) with good generalization capability is a complex process especially when the weights are severely quantized. Model averaging is a promising approach for achieving the good generalization capability of DNNs, especially when the loss surface for training contains many sharp minima. We present a new quantized neural network optimization approach, stochastic quantized weight averaging (SQWA), to design low-precision DNNs with good generalization capability using model averaging. The proposed approach includes (1) floating-point model training, (2) direct quantization of weights, (3) capturing multiple low-precision models during retraining with cyclical learning rates, (4) averaging the captured models, and (5) re-quantizing the averaged model and fine-tuning it with low-learning rates. Additionally, we present a loss-visualization technique on the quantized weight domain to clearly elucidate the behavior of the proposed method. Visualization results indicate that a quantized DNN (QDNN) optimized with the proposed approach is located near the center of the flat minimum in the loss surface. With SQWA training, we achieved state-of-the-art results for 2-bit QDNNs on CIFAR-100 and ImageNet datasets. Although we only employed a uniform quantization scheme for the sake of implementation in VLSI or low-precision neural processing units, the performance achieved exceeded those of previous studies employing non-uniform quantization.

Sungho Shin, Yoonho Boo, Wonyong Sung

Knowledge distillation (KD) is a very popular method for model size reduction. Recently, the technique is exploited for quantized deep neural networks (QDNNs) training as a way to restore the performance sacrificed by word-length reduction. KD, however, employs additional hyper-parameters, such as temperature, coefficient, and the size of teacher network for QDNN training. We analyze the effect of these hyper-parameters for QDNN optimization with KD. We find that these hyper-parameters are inter-related, and also introduce a simple and effective technique that reduces \textit{coefficient} during training. With KD employing the proposed hyper-parameters, we achieve the test accuracy of 92.7% and 67.0% on Resnet20 with 2-bit ternary weights for CIFAR-10 and CIFAR-100 data sets, respectively.

Wonyong Sung, Jinhwan Park

As neural network algorithms show high performance in many applications, their efficient inference on mobile and embedded systems are of great interests. When a single stream recurrent neural network (RNN) is executed for a personal user in embedded systems, it demands a large amount of DRAM accesses because the network size is usually much bigger than the cache size and the weights of an RNN are used only once at each time step. We overcome this problem by parallelizing the algorithm and executing it multiple time steps at a time. This approach also reduces the power consumption by lowering the number of DRAM accesses. QRNN (Quasi Recurrent Neural Networks) and SRU (Simple Recurrent Unit) based recurrent neural networks are used for implementation. The experiments for SRU showed about 300% and 930% of speed-up when the numbers of multi time steps are 4 and 16, respectively, in an ARM CPU based system.

Yoonho Boo, Wonyong Sung

Deep neural networks (DNNs) usually demand a large amount of operations for real-time inference. Especially, fully-connected layers contain a large number of weights, thus they usually need many off-chip memory accesses for inference. We propose a weight compression method for deep neural networks, which allows values of +1 or -1 only at predetermined positions of the weights so that decoding using a table can be conducted easily. For example, the structured sparse (8,2) coding allows at most two non-zero values among eight weights. This method not only enables multiplication-free DNN implementations but also compresses the weight storage by up to x32 compared to floating-point networks. Weight distribution normalization and gradual pruning techniques are applied to mitigate the performance degradation. The experiments are conducted with fully-connected deep neural networks and convolutional neural networks.

Sungho Shin, Yoonho Boo, Wonyong Sung

Fixed-point optimization of deep neural networks plays an important role in hardware based design and low-power implementations. Many deep neural networks show fairly good performance even with 2- or 3-bit precision when quantized weights are fine-tuned by retraining. We propose an improved fixedpoint optimization algorithm that estimates the quantization step size dynamically during the retraining. In addition, a gradual quantization scheme is also tested, which sequentially applies fixed-point optimizations from high- to low-precision. The experiments are conducted for feed-forward deep neural networks (FFDNNs), convolutional neural networks (CNNs), and recurrent neural networks (RNNs).

Sungho Shin, Kyuyeon Hwang, Wonyong Sung

The complexity of deep neural network algorithms for hardware implementation can be lowered either by scaling the number of units or reducing the word-length of weights. Both approaches, however, can accompany the performance degradation although many types of research are conducted to relieve this problem. Thus, it is an important question which one, between the network size scaling and the weight quantization, is more effective for hardware optimization. For this study, the performances of fully-connected deep neural networks (FCDNNs) and convolutional neural networks (CNNs) are evaluated while changing the network complexity and the word-length of weights. Based on these experiments, we present the effective compression ratio (ECR) to guide the trade-off between the network size and the precision of weights when the hardware resource is limited.

Sajid Anwar, Wonyong Sung

The learning capability of a neural network improves with increasing depth at higher computational costs. Wider layers with dense kernel connectivity patterns furhter increase this cost and may hinder real-time inference. We propose feature map and kernel level pruning for reducing the computational complexity of a deep convolutional neural network. Pruning feature maps reduces the width of a layer and hence does not need any sparse representation. Further, kernel pruning converts the dense connectivity pattern into a sparse one. Due to coarse nature, these pruning granularities can be exploited by GPUs and VLSI based implementations. We propose a simple and generic strategy to choose the least adversarial pruning masks for both granularities. The pruned networks are retrained which compensates the loss in accuracy. We obtain the best pruning ratios when we prune a network with both granularities. Experiments with the CIFAR-10 dataset show that more than 85% sparsity can be induced in the convolution layers with less than 1% increase in the missclassification rate of the baseline network.

Minjae Lee, Kyuyeon Hwang, Jinhwan Park et al.

In this paper, a neural network based real-time speech recognition (SR) system is developed using an FPGA for very low-power operation. The implemented system employs two recurrent neural networks (RNNs); one is a speech-to-character RNN for acoustic modeling (AM) and the other is for character-level language modeling (LM). The system also employs a statistical word-level LM to improve the recognition accuracy. The results of the AM, the character-level LM, and the word-level LM are combined using a fairly simple N-best search algorithm instead of the hidden Markov model (HMM) based network. The RNNs are implemented using massively parallel processing elements (PEs) for low latency and high throughput. The weights are quantized to 6 bits to store all of them in the on-chip memory of an FPGA. The proposed algorithm is implemented on a Xilinx XC7Z045, and the system can operate much faster than real-time.

Kyuyeon Hwang, Wonyong Sung

Recurrent neural network (RNN) based character-level language models (CLMs) are extremely useful for modeling out-of-vocabulary words by nature. However, their performance is generally much worse than the word-level language models (WLMs), since CLMs need to consider longer history of tokens to properly predict the next one. We address this problem by proposing hierarchical RNN architectures, which consist of multiple modules with different timescales. Despite the multi-timescale structures, the input and output layers operate with the character-level clock, which allows the existing RNN CLM training approaches to be directly applicable without any modifications. Our CLM models show better perplexity than Kneser-Ney (KN) 5-gram WLMs on the One Billion Word Benchmark with only 2% of parameters. Also, we present real-time character-level end-to-end speech recognition examples on the Wall Street Journal (WSJ) corpus, where replacing traditional mono-clock RNN CLMs with the proposed models results in better recognition accuracies even though the number of parameters are reduced to 30%.

Sungho Shin, Wonyong Sung

Gesture recognition is a very essential technology for many wearable devices. While previous algorithms are mostly based on statistical methods including the hidden Markov model, we develop two dynamic hand gesture recognition techniques using low complexity recurrent neural network (RNN) algorithms. One is based on video signal and employs a combined structure of a convolutional neural network (CNN) and an RNN. The other uses accelerometer data and only requires an RNN. Fixed-point optimization that quantizes most of the weights into two bits is conducted to optimize the amount of memory size for weight storage and reduce the power consumption in hardware and software based implementations.

Sungho Shin, Kyuyeon Hwang, Wonyong Sung

In this paper, we propose a generative knowledge transfer technique that trains an RNN based language model (student network) using text and output probabilities generated from a previously trained RNN (teacher network). The text generation can be conducted by either the teacher or the student network. We can also improve the performance by taking the ensemble of soft labels obtained from multiple teacher networks. This method can be used for privacy conscious language model adaptation because no user data is directly used for training. Especially, when the soft labels of multiple devices are aggregated via a trusted third party, we can expect very strong privacy protection.

Jinhwan Park, Wonyong Sung

Deep neural networks (DNNs) demand a very large amount of computation and weight storage, and thus efficient implementation using special purpose hardware is highly desired. In this work, we have developed an FPGA based fixed-point DNN system using only on-chip memory not to access external DRAM. The execution time and energy consumption of the developed system is compared with a GPU based implementation. Since the capacity of memory in FPGA is limited, only 3-bit weights are used for this implementation, and training based fixed-point weight optimization is employed. The implementation using Xilinx XC7Z045 is tested for the MNIST handwritten digit recognition benchmark and a phoneme recognition task on TIMIT corpus. The obtained speed is about one quarter of a GPU based implementation and much better than that of a PC based one. The power consumption is less than 5 Watt at the full speed operation resulting in much higher efficiency compared to GPU based systems.

Kyuyeon Hwang, Wonyong Sung

In real-time speech recognition applications, the latency is an important issue. We have developed a character-level incremental speech recognition (ISR) system that responds quickly even during the speech, where the hypotheses are gradually improved while the speaking proceeds. The algorithm employs a speech-to-character unidirectional recurrent neural network (RNN), which is end-to-end trained with connectionist temporal classification (CTC), and an RNN-based character-level language model (LM). The output values of the CTC-trained RNN are character-level probabilities, which are processed by beam search decoding. The RNN LM augments the decoding by providing long-term dependency information. We propose tree-based online beam search with additional depth-pruning, which enables the system to process infinitely long input speech with low latency. This system not only responds quickly on speech but also can dictate out-of-vocabulary (OOV) words according to pronunciation. The proposed model achieves the word error rate (WER) of 8.90% on the Wall Street Journal (WSJ) Nov'92 20K evaluation set when trained on the WSJ SI-284 training set.

Kyuyeon Hwang, Minjae Lee, Wonyong Sung

In this paper, we propose a context-aware keyword spotting model employing a character-level recurrent neural network (RNN) for spoken term detection in continuous speech. The RNN is end-to-end trained with connectionist temporal classification (CTC) to generate the probabilities of character and word-boundary labels. There is no need for the phonetic transcription, senone modeling, or system dictionary in training and testing. Also, keywords can easily be added and modified by editing the text based keyword list without retraining the RNN. Moreover, the unidirectional RNN processes an infinitely long input audio streams without pre-segmentation and keywords are detected with low-latency before the utterance is finished. Experimental results show that the proposed keyword spotter significantly outperforms the deep neural network (DNN) and hidden Markov model (HMM) based keyword-filler model even with less computations.

Sajid Anwar, Kyuyeon Hwang, Wonyong Sung

Real time application of deep learning algorithms is often hindered by high computational complexity and frequent memory accesses. Network pruning is a promising technique to solve this problem. However, pruning usually results in irregular network connections that not only demand extra representation efforts but also do not fit well on parallel computation. We introduce structured sparsity at various scales for convolutional neural networks, which are channel wise, kernel wise and intra kernel strided sparsity. This structured sparsity is very advantageous for direct computational resource savings on embedded computers, parallel computing environments and hardware based systems. To decide the importance of network connections and paths, the proposed method uses a particle filtering approach. The importance weight of each particle is assigned by computing the misclassification rate with corresponding connectivity pattern. The pruned network is re-trained to compensate for the losses due to pruning. While implementing convolutions as matrix products, we particularly show that intra kernel strided sparsity with a simple constraint can significantly reduce the size of kernel and feature map matrices. The pruned network is finally fixed point optimized with reduced word length precision. This results in significant reduction in the total storage size providing advantages for on-chip memory based implementations of deep neural networks.

Sungho Shin, Kyuyeon Hwang, Wonyong Sung

Recurrent neural networks have shown excellent performance in many applications, however they require increased complexity in hardware or software based implementations. The hardware complexity can be much lowered by minimizing the word-length of weights and signals. This work analyzes the fixed-point performance of recurrent neural networks using a retrain based quantization method. The quantization sensitivity of each layer in RNNs is studied, and the overall fixed-point optimization results minimizing the capacity of weights while not sacrificing the performance are presented. A language model and a phoneme recognition examples are used.

Kyuyeon Hwang, Wonyong Sung

Connectionist temporal classification (CTC) based supervised sequence training of recurrent neural networks (RNNs) has shown great success in many machine learning areas including end-to-end speech and handwritten character recognition. For the CTC training, however, it is required to unroll (or unfold) the RNN by the length of an input sequence. This unrolling requires a lot of memory and hinders a small footprint implementation of online learning or adaptation. Furthermore, the length of training sequences is usually not uniform, which makes parallel training with multiple sequences inefficient on shared memory models such as graphics processing units (GPUs). In this work, we introduce an expectation-maximization (EM) based online CTC algorithm that enables unidirectional RNNs to learn sequences that are longer than the amount of unrolling. The RNNs can also be trained to process an infinitely long input sequence without pre-segmentation or external reset. Moreover, the proposed approach allows efficient parallel training on GPUs. For evaluation, phoneme recognition and end-to-end speech recognition examples are presented on the TIMIT and Wall Street Journal (WSJ) corpora, respectively. Our online model achieves 20.7% phoneme error rate (PER) on the very long input sequence that is generated by concatenating all 192 utterances in the TIMIT core test set. On WSJ, a network can be trained with only 64 times of unrolling while sacrificing 4.5% relative word error rate (WER).

Wonyong Sung, Sungho Shin, Kyuyeon Hwang

The complexity of deep neural network algorithms for hardware implementation can be much lowered by optimizing the word-length of weights and signals. Direct quantization of floating-point weights, however, does not show good performance when the number of bits assigned is small. Retraining of quantized networks has been developed to relieve this problem. In this work, the effects of retraining are analyzed for a feedforward deep neural network (FFDNN) and a convolutional neural network (CNN). The network complexity is controlled to know their effects on the resiliency of quantized networks by retraining. The complexity of the FFDNN is controlled by varying the unit size in each hidden layer and the number of layers, while that of the CNN is done by modifying the feature map configuration. We find that the performance gap between the floating-point and the retrain-based ternary (+1, 0, -1) weight neural networks exists with a fair amount in 'complexity limited' networks, but the discrepancy almost vanishes in fully complex networks whose capability is limited by the training data, rather than by the number of connections. This research shows that highly complex DNNs have the capability of absorbing the effects of severe weight quantization through retraining, but connection limited networks are less resilient. This paper also presents the effective compression ratio to guide the trade-off between the network size and the precision when the hardware resource is limited.

Kyuyeon Hwang, Wonyong Sung

Recurrent neural networks (RNNs) have shown outstanding performance on processing sequence data. However, they suffer from long training time, which demands parallel implementations of the training procedure. Parallelization of the training algorithms for RNNs are very challenging because internal recurrent paths form dependencies between two different time frames. In this paper, we first propose a generalized graph-based RNN structure that covers the most popular long short-term memory (LSTM) network. Then, we present a parallelization approach that automatically explores parallelisms of arbitrary RNNs by analyzing the graph structure. The experimental results show that the proposed approach shows great speed-up even with a single training stream, and further accelerates the training when combined with multiple parallel training streams.