Universally Quantized Neural Compression

• URL: http://arxiv.org/abs/2006.09952v2
• Date: Wed, 21 Oct 2020 17:10:20 GMT
• Title: Universally Quantized Neural Compression
• Abstract summary: We show that a uniform noise channel can be implemented at test time using universal quantization. This allows us to eliminate the mismatch between training and test phases while maintaining a completely differentiable loss function.
• Score: 25.808505103655687
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: A popular approach to learning encoders for lossy compression is to use additive uniform noise during training as a differentiable approximation to test-time quantization. We demonstrate that a uniform noise channel can also be implemented at test time using universal quantization (Ziv, 1985). This allows us to eliminate the mismatch between training and test phases while maintaining a completely differentiable loss function. Implementing the uniform noise channel is a special case of the more general problem of communicating a sample, which we prove is computationally hard if we do not make assumptions about its distribution. However, the uniform special case is efficient as well as easy to implement and thus of great interest from a practical point of view. Finally, we show that quantization can be obtained as a limiting case of a soft quantizer applied to the uniform noise channel, bridging compression with and without quantization.

Related papers

• Sampling (noisy) quantum circuits through randomized rounding [6.09006146869448]
  We show how to replicate noisy quantum circuits for optimisation problems.<n>We run large-scale simulations and experiments on IBMQ hardware.<n>We provide a benchmark for future error-mitigated or fault-tolerant demonstrations of quantum advantage.
  arXiv  Detail & Related papers  (2025-07-29T14:56:17Z)
• Optimizing Learned Image Compression on Scalar and Entropy-Constraint Quantization [8.95146413290727]
  We show that retraining on correctly quantized data consistently yields additional coding gain for both uniform scalar and especially for entropy-constraint quantization.<n>For the Kodak test set, we obtain average savings between 1% and 2%, and for the TecNick test set up to 2.2% in terms of Bjontegaard-Delta.
  arXiv  Detail & Related papers  (2025-06-10T10:22:22Z)
• Quantum computational advantage of noisy boson sampling with partially distinguishable photons [0.0]
  We identify the level of partial distinguishability noise that upholds the classical intractability of boson sampling. We find that boson sampling with on average $O(log N)$ number of distinguishable photons out of $N$ input photons maintains the equivalent complexity to the ideal boson sampling case.
  arXiv  Detail & Related papers  (2025-01-23T07:37:29Z)
• Classical Verification of Quantum Learning Advantages with Noises [0.27930367518472443]
  We propose an efficient classical error rectification algorithm to reconstruct the noise-free results given by the quantum Fourier sampling circuit. We also prove that a classical client with access to the random example oracle can verify the agnostic parity learning results from the noisy quantum prover.
  arXiv  Detail & Related papers  (2024-11-14T06:14:39Z)
• Entanglement-enabled advantage for learning a bosonic random displacement channel [6.354564715928202]
  We show that quantum entanglement can provide an exponential advantage in learning properties of a bosonic continuous-variable system. We present a simple entanglement-assisted scheme that only requires a number of samples independent of $n$, given a sufficient amount of squeezing. Our work illuminates the role of entanglement in learning continuous-variable systems and points toward experimentally feasible demonstrations of provable entanglement-enabled advantage.
  arXiv  Detail & Related papers  (2024-02-29T02:30:39Z)
• Compressed quantum error mitigation [0.0]
  We introduce a quantum error mitigation technique based on probabilistic error cancellation to eliminate errors which have accumulated during the application of a quantum circuit. For a simple noise model, we show that efficient, local denoisers can be found, and we demonstrate their effectiveness for the digital quantum simulation of the time evolution of simple spin chains.
  arXiv  Detail & Related papers  (2023-02-10T19:00:02Z)
• Importance sampling for stochastic quantum simulations [68.8204255655161]
  We introduce the qDrift protocol, which builds random product formulas by sampling from the Hamiltonian according to the coefficients. We show that the simulation cost can be reduced while achieving the same accuracy, by considering the individual simulation cost during the sampling stage. Results are confirmed by numerical simulations performed on a lattice nuclear effective field theory.
  arXiv  Detail & Related papers  (2022-12-12T15:06:32Z)
• Suppressing Amplitude Damping in Trapped Ions: Discrete Weak Measurements for a Non-unitary Probabilistic Noise Filter [62.997667081978825]
  We introduce a low-overhead protocol to reverse this degradation. We present two trapped-ion schemes for the implementation of a non-unitary probabilistic filter against amplitude damping noise. This filter can be understood as a protocol for single-copy quasi-distillation.
  arXiv  Detail & Related papers  (2022-09-06T18:18:41Z)
• Unified Multivariate Gaussian Mixture for Efficient Neural Image Compression [151.3826781154146]
  latent variables with priors and hyperpriors is an essential problem in variational image compression. We find inter-correlations and intra-correlations exist when observing latent variables in a vectorized perspective. Our model has better rate-distortion performance and an impressive $3.18times$ compression speed up.
  arXiv  Detail & Related papers  (2022-03-21T11:44:17Z)
• Noise thresholds for classical simulability of non-linear Boson sampling [4.812718493682455]
  We introduce higher order non-linearities as a mean to enhance the computational complexity of the problem and the protocol's robustness against noise. Our results indicate that the addition of single-mode Kerr non-linearity at the input state preparation level, while retaining a linear-optical evolution, makes the Boson sampling protocol more robust against noise.
  arXiv  Detail & Related papers  (2022-02-24T12:17:28Z)
• Tight bounds on the convergence of noisy random circuits to the uniform distribution [1.4841630983274847]
  We study the properties of output distributions of noisy, random circuits. We discuss recent barrier results for depth-agnostic and/or noise-agnostic proof techniques.
  arXiv  Detail & Related papers  (2021-12-01T18:47:50Z)
• Nonuniform-to-Uniform Quantization: Towards Accurate Quantization via Generalized Straight-Through Estimation [48.838691414561694]
  Nonuniform-to-Uniform Quantization (N2UQ) is a method that can maintain the strong representation ability of nonuniform methods while being hardware-friendly and efficient. N2UQ outperforms state-of-the-art nonuniform quantization methods by 0.71.8% on ImageNet.
  arXiv  Detail & Related papers  (2021-11-29T18:59:55Z)
• Forster Decomposition and Learning Halfspaces with Noise [60.691817861402676]
  A Forster transform is an operation that turns a distribution into one with good anti-concentration properties. We show that any distribution can be decomposed efficiently as a disjoint mixture of few distributions for which a Forster transform exists and can be computed efficiently.
  arXiv  Detail & Related papers  (2021-07-12T17:00:59Z)
• Sigma-Delta and Distributed Noise-Shaping Quantization Methods for Random Fourier Features [73.25551965751603]
  We prove that our quantized RFFs allow a high accuracy approximation of the underlying kernels. We show that the quantized RFFs can be further compressed, yielding an excellent trade-off between memory use and accuracy. We empirically show by testing the performance of our methods on several machine learning tasks that our method compares favorably to other state of the art quantization methods in this context.
  arXiv  Detail & Related papers  (2021-06-04T17:24:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.