Current compression systems incorporate a data model, however formed, deeply into the coding process, leading to difficulties of an architectural nature. This work contributes an alternative "Model-Code Separation" architecture for general compression, based on model-free coding and iterative message-passing algorithms over graphical models representing the modeling and coding aspects of compression in decoding. Systems following this architecture resolve important challenges posed by current systems, and stand to benefit further from advances in the understanding of data and the algorithms that process them.

High-performance Model-Code Separation (MCS) architectures for lossless compression are practically viable with graphical message-passing in the decoder. This paper extends separation architectures to lossy compression by constructing model-free but semantics-aware encoders and contributes a new inference-friendly low-density hashing quantizer (LDHQ) to support decoding.

As cloud computing is taking off, the presence of high-performance interactive Internet applications is exploding. By nature, these applications require responsive client-server data exchange and lossless, in-order delivery. Previous work has shown that by using forward error correction (FEC), it is possible to reduce the data streaming latency caused by retransmissions of lost packets. However, the prior schemes only send FEC packets when there are no original packets pending transmission. In this paper, we further expand the hybrid FEC-ARQ protocol and show that sometimes, the transmission latency can be further reduced by preempting original data packets with FEC packets. We have formulated the decision of whether to send new original data packets, FEC packets, or resend original data packets as a transmission policy. An optimal transmission policy is selected to minimize the delay experienced by the application subject to a constraint on the amount of overhead. By using this optimal policy, we significantly improve the delay performance over straightforward FEC schemes while controlling the amount of overhead due to FEC.

We propose a linear predictive quantization system for causally transmitting parallel sources with temporal memory (colored frames) over an erasure channel. By optimizing within this structure, we derive an achievability result in the high-rate limit and compare it to an upper bound on performance. The proposed system subsumes the well-known PCM and DPCM systems as special cases. While typically DPCM performs well without erasures and PCM suffers less with many erasures, we show that the proposed solution improves performance over both under all severities of erasures, with unbounded improvement in some cases.

There is an ongoing explosion of interactive Internet applications. By nature, these applications require responsive client-server data exchange and lossless, in-order delivery. In previous work, we have shown that a hybrid FEC-ARQ protocol which combines sending original data packets with forward error correction (FEC) packets is effective in reducing the latency caused by retransmissions of lost packets. However, the prior scheme only sends FEC packets when there are no original packets pending transmission. In this paper, we further expand the investigation of the hybrid FEC-ARQ protocol and show that sometimes, the transmission latency can be further reduced by preempting original data packets with FEC packets. We have formulated the decision of whether to send new original data packets, FEC packets, or resend original data packets as a transmission policy. An optimal transmission policy is selected to minimize the delay experienced by the application subject to a constraint on the amount of waste. By using this optimal policy, we significantly improve the delay performance over straightforward FEC schemes while controlling the amount of waste due to FEC.

Interactive Internet applications that rely on sequential streams for lossless data exchange often use retransmission protocols (e.g. TCP) for reliability and the guarantee of sequential data ordering. More so than for bulk file transfer or media delivery, lossless sequential streaming poses an even greater challenge for the common problem cases of retransmission protocols, such as lossy links or long network paths, manifesting as significant latency in the interactive user experience. We propose a hybrid FEC-ARQ protocol built on a packet streaming code that reduces to a simple strategy over sending or resending original data packets or check packets combining undecoded packets, based on actual network conditions. Experimental results show that our proposed protocol can significantly improve the total delay over retransmission and other schemes that use FEC, under a range of bandwidth and loss scenarios.

Media delivery over packet networks is often plagued by packet losses which limit its utility to end users. Forward error correction (FEC) based techniques are important for overcoming this problem. This paper further develops an FEC-based technique to maximize the expected received media quality by jointly choosing which packets to send and which packets to protect - including discarding packets to make additional room for protection. We describe a straight-forward implementation leveraging existing FEC system components. Comprehensive experiments demonstrate that significant gains in PSNR of several dB are achieved when sending H.264/MPEG-4 AVC coded video over a packet erasure channel.

Media delivery over lossy packet networks is a challenging problem, and forward error correction (FEC) based techniques are an important technique for overcoming packet loss. Conventional FEC-based media delivery techniques protect all packets equally, or protect a subset of the packets, or protect different subsets of packets with different levels of protection, e.g., scalable coding with unequal error protection (UEP). This paper proposes an FEC-based technique to maximize the expected received media quality by explicitly discarding packets, when beneficial, in order to provide additional room for FEC. Given knowledge of the importance of each packet, we show that there is a simple and intuitive criterion for the optimal selection of which packets to discard and which to protect, as well as the level of protection, to minimize the expected distortion experienced at the receiver. The proposed approach provides significant gains over the conventional approaches, and these gains are illustrated for the case of sending H.264 coded video data over a packet erasure channel with known packet loss rate.

Gauss mixture (GM) models are frequently used for their ability to well approximate many densities and for their tractability to analysis. We propose new classification methods built on GM clustering algorithms more often studied and used for vector quantization (VQ). One of our methods is an extension of the 'codebook matching' idea to the specific case of classifying whole images. We apply these methods to a realistic supervised classification problem and empirically evaluate their performances compared with other classification methods.