The KO (Kronecker Operation) code is a recent deep-learned error-correcting code using a neural network architecture to generalize a Reed-Muller code with Dumer decoding. Analyzing the encoder modules and using ablation techniques, we give interpretations of the KO encoder which significantly reduce the number of parameters. We also discuss interpretability aspects of the KO decoder. The interpretation opens up possibilities to give an explicit representation of KO codes, which could be useful for more efficient learning of KO codes and explaining the learning mechanism underlying the empirical observations made about its performance in previous work.
The KO (Kronecker Operation) code is a recent deep-learned error-correcting code using a neural network architecture to generalize a Reed-Muller code with Dumer decoding. Analyzing the encoder modules and using ablation techniques, we give interpretations of the KO encoder which significantly reduce the number of parameters. We also discuss interpretability aspects of the KO decoder. The interpretation opens up possibilities to give an explicit representation of KO codes, which could be useful for more efficient learning of KO codes and explaining the learning mechanism underlying the empirical observations made about its performance in previous work.
Codes are the backbone of modern information age.Codes, composed of encoder and decoder pairs, are the basic mathematical objects that enable reliable communication.Landmark codes include convolutional, Reed-Muller, turbo, LDPC, and polar: each is linear and represents a mathematical breakthrough.Their impact on humanity is huge; each of these codes has been used in global communication standards over the past six decades.On the other hand, designing codes is a challenging task, mostly driven by human ingenuity.Befittingly, historically, the progress in discovery of codes has been sporadic.In this thesis, we present a new paradigm to invent codes via harnessing tools from deep-learning.Our major result is the invention of KO codes, a computationally efficient family of deep-learning driven codes that outperform the state-of-the-art RM and polar codes, in the challenging short-to-medium block length regime.The key technical innovation behind KO codes is the design of a novel family of neural architectures inspired by the computation tree of the Kronecker Operation (KO) central to RM and polar codes.These architectures pave the way for discovery of a much richer class of hitherto unexplored non-linear codes.This design technique can be viewed as an instantiation of the classical neural augmentation principle.In the process, we also study a popular neural network model called Mixture-of-Experts (MoE) that realizes this principle.We provide the first set of consistent and efficient algorithms with global learning guarantees for learning the parameters in a MoE which has been an open problem for more than two decades.B.1 Comparison of SGD on our losses (L 4 , L log ) vs. ℓ 2 and the EM algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . .106 C.2 KO code achieves a significant gain over the Polar(64, 7) code in BLER when trained on AWGN channel.KO decoder also matches the optimal MAP decoder. . . . . . . . . . . . .108 xii C.6 KO(6, 1) code.Left: KO(6, 1) code achieves significant gain over RM(6, 1) code (with Dumer) when trained on AWGN channel.Right: Under the optimal MAP decoding, KO(6, 1) and RM(6, 1) codes achieve the same performance.Error rates for a random Gaussian codebook are also plotted as a baseline. . . . . . . . . . . . . . . . . . . . . . . . . . . .120 C.1 Plotkin trees for the Polar(64, 7) encoder and our neural KO encoder counterpart.Both codes have dimension k = 7 and blocklength n = 64. . . . . . . . . . . . . . . . . . . . . .121 C.3 The same KO(8,2) encoder and decoder as those used in Figure 4.4 achieve a significant gain (without any retraining or fine-tuning) when list decoding is used together with the KO decoder.The magnitude of the gain is comparable to the gain achieved by the same list decoding technique on the successive cancellation decoder of the RM(8,2) code.We used the list decoding from [1] but without the permutation technique. . . . . . . . . . . . . . . . . . . . . . . . . . . . .122 C.4 Plotkin trees for RM(8,2) and KO(8,2) encoders.Leaves are highlighted in green.Both codes have dimension k = 37 and blocklength n = 256. . . . . . . . . . . . . . . . . . . . . .123 C.5 Plotkin trees for the RM(8, 2) and KO(8, 2) decoders.Red arrows indicate the bit decoding order. . . . . . . .
This paper introduces HiKO (Hierarchical Kronecker Operation), a novel framework for training high-rate neural error-correcting codes that enables KO codes to outperform Reed-Muller codes beyond second order. To our knowledge, this is the first attempt to extend KO codes beyond second order. While conventional KO codes show promising results for low-rate regimes (r<2), they degrade at higher rates -- a critical limitation for practical deployment. Our framework incorporates three key innovations: (1) a hierarchical training methodology that decomposes complex high-rate codes into simpler constituent codes for efficient knowledge transfer, (2) enhanced neural architectures with dropout regularization and learnable skip connections tailored for the Plotkin structure, and (3) a progressive unfreezing strategy that systematically transitions from pre-trained components to fully optimized integrated codes. Our experiments show that HiKO codes consistently outperform traditional Reed-Muller codes across various configurations, achieving notable performance improvements for third-order (r=3) and fourth-order (r=4) codes. Analysis reveals that HiKO codes successfully approximate Shannon-optimal Gaussian codebooks while preserving efficient decoding properties. This represents the first successful extension of KO codes beyond second order, opening new possibilities for neural code deployment in high-throughput communication systems.
As new deep-learned error-correcting codes continue to be introduced, it is important to develop tools to interpret the designed codes and understand the training process. Prior work focusing on the deep-learned TurboAE has both interpreted the learned encoders post-hoc by mapping these onto nearby "interpretable" encoders, and experimentally evaluated the performance of these interpretable encoders with various decoders. Here we look at developing tools for interpreting the training process for deep-learned error-correcting codes, focusing on: 1) using the Goldreich-Levin algorithm to quickly interpret the learned encoder; 2) using Fourier coefficients as a tool for understanding the training dynamics and the loss landscape; 3) reformulating the training loss, the binary cross entropy, by relating it to encoder and decoder parameters, and the bit error rate (BER); 4) using these insights to formulate and study a new training procedure. All tools are demonstrated on TurboAE, but are applicable to other deep-learned forward error correcting codes (without feedback).
We revisit the idea of using deep neural networks for one-shot decoding of random and structured codes, such as polar codes. Although it is possible to achieve maximum a posteriori (MAP) bit error rate (BER) performance for both code families and for short codeword lengths, we observe that (i) structured codes are easier to learn and (ii) the neural network is able to generalize to codewords that it has never seen during training for structured, but not for random codes. These results provide some evidence that neural networks can learn a form of decoding algorithm, rather than only a simple classifier. We introduce the metric normalized validation error (NVE) in order to further investigate the potential and limitations of deep learning-based decoding with respect to performance and complexity.
We revisit the idea of using deep neural networks for one-shot decoding of random and structured codes, such as polar codes. Although it is possible to achieve maximum a posteriori (MAP) bit error rate (BER) performance for both code families and for short codeword lengths, we observe that (i) structured codes are easier to learn and (ii) the neural network is able to generalize to codewords that it has never seen during training for structured, but not for random codes. These results provide some evidence that neural networks can learn a form of decoding algorithm, rather than only a simple classifier. We introduce the metric normalized validation error (NVE) in order to further investigate the potential and limitations of deep learning-based decoding with respect to performance and complexity.
Abstract Neural‐network (NN) based decoders are becoming increasingly popular in the field of quantum error correction (QEC), including for decoding of state‐of‐the‐art quantum computation experiments. In this work, established interpretability methods are used from the field of machine learning, to introduce a toolbox to achieve an understanding of the underlying decoding logic of NN decoders, which have been trained but otherwise typically operate as black‐box models. To illustrate the capabilities of the employed interpretability method, based on the Shapley value approximation, an examplary case study of a NN decoder is provided that is trained for flag‐qubit based fault‐tolerant (FT) QEC with the Steane code. The interpretation of particular decoding decisions of the NN is analysed, by doing so it is revealed how the NN learns to capture fundamental structures in the information gained from syndrome and flag qubit measurements, in order to come to a FT correction decision. Further, it is showed that the understanding of how the NN obtains a decoding decision can be used on the one hand to identify flawed processing of error syndrome information by the NN, resulting in decreased decoding performance, as well as for well‐informed improvements of the NN architecture. The diagnostic capabilities of the interpretability method presented here can help ensure successful application of machine learning for decoding of QEC protocols.