Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

“Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks”の学習報告

Digital processing of speech signals

An intelligent automated assistant system engages with the user in an integrated, conversational manner using natural language dialog, and invokes external services when appropriate to obtain information or perform various actions. The system can be implemented using any of a number of different platforms, such as the web, email, smartphone, and the like, or any combination thereof. In one embodiment, the system is based on sets of interrelated domains and tasks, and employs additional functionally powered by external services with which the system can interact.

Intelligent Automated Assistant

When building a unified vision system or gradually adding new capabilities to a system, the usual assumption is that training data for all tasks is always available. However, as the number of tasks grows, storing and retraining on such data becomes infeasible. A new problem arises where we add new capabilities to a Convolutional Neural Network (CNN), but the training data for its existing capabilities are unavailable. We propose our Learning without Forgetting method, which uses only new task data to train the network while preserving the original capabilities. Our method performs favorably compared to commonly used feature extraction and fine-tuning adaption techniques and performs similarly to multitask learning that uses original task data we assume unavailable. A more surprising observation is that Learning without Forgetting may be able to replace fine-tuning with similar old and new task datasets for improved new task performance.

Learning without Forgetting

| Convolutional coding as a means oferror correction is widely used in mobile communi-cations, where the effects of channel noise have to betaken into account. Channel decoders usually aim atminimizing the frame, symbol or residual bit errorrate of the source bits.In this contribution we show how the residual bit er-ror rate can be reduced by using a priori knowledgeabout the source statistics within the channel decod-ing process. This usually helps the source decoderto perform better. If the source decoder also usesa priori knowledge, however, we show that the bestoverall system performance is reached at a higherresidual bit error rate after channel decoding. Thisgives new insight into optimization criteria of com-bined source/channel decoding.I. I NTRODUCTION Real world source coding usually leaves a certainamount of residual redundancy within the bit stream.This is mostly due to constraints on coding delay andcomplexity. In recent years an increasing interest hasgrown in decoding techniques that are able to make useof this residual source redundancy.Bahl et al. laid the foundations of symbol-by-symbolchannel decoding that is able to exploit a priori knowl-edge about the source bits [1]. Hagenauer investigatedsequence-estimating channel decoding algorithms andproposed the source-controlled channel decodingtech-nique [2]. Further work focused on channel decodingexploiting source statistics to reduce the residual path,symbol, or bit error rate is [3–5].Alternatively, the source redundancy can of course beused in the source decoding process [6–14] often re-ferred to as soft decision or softbit decoding. In contrastto channeldecodingthese techniqueshavethe advantageof an easy adaptation to arbitrary error criteria, e.g. thesignal-to-noise ratio rather than the bit error rate.In this paper we investigate combined source/channeldecoding in the sense that the channel decoder as wellas the source decoder are able to exploit source statis-tics. Furthermore, the channel decoder yields soft deci-sions (softbits) for use in the softbit source decoder. Insection II we brieﬂy review the softbit decoding tech-nique [11,12], whereas in section III we discuss sourcecontrolled channel decoding according to [5,15], bothpresented in a uniform manner. Simulation results willbe presented answering the question, whether a prioriknowledge can be used twice, in the channel decoder

Combined Source/Channel Decoding: When Minimizing Bit Error Rate is Suboptimal

Scalar Quantization With Optimized Receiver-Sided Adaptive Codebook Reconstruction Levels Controlled by a Predictor

DNN/CNN Acoustic Model Turbo Fusion for Phoneme Recognition.

A 360° perception of scene geometry is essential for automated driving, notably for parking and urban driving scenarios. Typically, it is achieved using surround-view fisheye cameras, focusing on the near-field area around the vehicle. The majority of current depth estimation approaches focus on employing just a single camera, which cannot be straightforwardly generalized to multiple cameras. The depth estimation model must be tested on a variety of cameras equipped to millions of cars with varying camera geometries. Even within a single car, intrinsics vary due to manufacturing tolerances. Deep learning models are sensitive to these changes, and it is practically infeasible to train and test on each camera variant. As a result, we present novel camera-geometry adaptive multi-scale convolutions which utilize the camera parameters as a conditional input, enabling the model to generalize to previously unseen fisheye cameras. Additionally, we improve the distance estimation by pairwise and patchwise vector-based self-attention encoder networks. We evaluate our approach on the Fisheye WoodScape surround-view dataset, significantly improving over previous approaches. We also show a generalization of our approach across different camera viewing angles and perform extensive experiments to support our contributions. To enable comparison with other approaches, we evaluate the front camera data on the KITTI dataset (pinhole camera images) and achieve state-of-the-art performance among self-supervised monocular methods. An overview video with qualitative results is provided at this https URL. Baseline code and dataset will be made public.

Tim Fingscheidt

Papers

Combined Source/Channel Decoding: When Minimizing Bit Error Rate is Suboptimal

Scalar Quantization With Optimized Receiver-Sided Adaptive Codebook Reconstruction Levels Controlled by a Predictor

DNN/CNN Acoustic Model Turbo Fusion for Phoneme Recognition.

SVDistNet: Self-Supervised Near-Field Distance Estimation on Surround View Fisheye Cameras

Multimodal ASR by Turbo Decoding vs. Feature Concatenation: Where to Perform Information Integration?