While the 2024 Nobel Prize in Physics ignites a worldwide discussion on the origins of neural networks and their foundational links to physics, modern machine learning research predominantly focuses on computational and algorithmic advancements, overlooking a picture of physics. Here we introduce the concept of entropy into neural networks by reconceptualizing them as hypothetical physical systems where each parameter is a non-interacting 'particle' within a one-dimensional space. By employing a Wang-Landau algorithms, we construct the neural networks' (with up to 1 million parameters) entropy landscapes as functions of training loss and test accuracy (or loss) across four distinct machine learning tasks, including arithmetic question, real-world tabular data, image recognition, and language modeling. Our results reveal the existence of \textit{entropy advantage}, where the high-entropy states generally outperform the states reached via classical training optimizer like stochastic gradient descent. We also find this advantage is more pronounced in narrower networks, indicating a need of different training optimizers tailored to different sizes of neural networks.

Reinforcement learning based recommender systems (RL-based RS) aim at learning a good policy from a batch of collected data, by casting recommendations to multi-step decision-making tasks. However, current RL-based RS research commonly has a large reality gap. In this paper, we introduce the first open-source real-world dataset, RL4RS, hoping to replace the artificial datasets and semi-simulated RS datasets previous studies used due to the resource limitation of the RL-based RS domain. Unlike academic RL research, RL-based RS suffers from the difficulties of being well-validated before deployment. We attempt to propose a new systematic evaluation framework, including evaluation of environment simulation, evaluation on environments, counterfactual policy evaluation, and evaluation on environments built from test set. In summary, the RL4RS (Reinforcement Learning for Recommender Systems), a new resource with special concerns on the reality gaps, contains two real-world datasets, data understanding tools, tuned simulation environments, related advanced RL baselines, batch RL baselines, and counterfactual policy evaluation algorithms. The RL4RS suite can be found at https://github.com/fuxiAIlab/RL4RS. In addition to the RL-based recommender systems, we expect the resource to contribute to research in applied reinforcement learning.

We propose a novel domain-specific generative pre-training (DS-GPT) method for text generation and apply it to the product titleand review summarization problems on E-commerce mobile display.First, we adopt a decoder-only transformer architecture, which fitswell for fine-tuning tasks by combining input and output all to-gether. Second, we demonstrate utilizing only small amount of pre-training data in related domains is powerful. Pre-training a languagemodel from a general corpus such as Wikipedia or the CommonCrawl requires tremendous time and resource commitment, andcan be wasteful if the downstream tasks are limited in variety. OurDSGPT is pre-trained on a limited dataset, the Chinese short textsummarization dataset (LCSTS). Third, our model does not requireproduct-related human-labeled data. For title summarization task,the state of art explicitly uses additional background knowledgein training and predicting stages. In contrast, our model implic-itly captures this knowledge and achieves significant improvementover other methods, after fine-tuning on the public Taobao.comdataset. For review summarization task, we utilize JD.com in-housedataset, and observe similar improvement over standard machinetranslation methods which lack the flexibility of fine-tuning. Ourproposed work can be simply extended to other domains for a widerange of text generation tasks.

Relevance has significant impact on user experience and business profit for e-commerce search platform. In this work, we propose a data-driven framework for search relevance prediction, by distilling knowledge from BERT and related multi-layer Transformer teacher models into simple feed-forward networks with large amount of unlabeled data. The distillation process produces a student model that recovers more than 97\% test accuracy of teacher models on new queries, at a serving cost that's several magnitude lower (latency 150x lower than BERT-Base and 15x lower than the most efficient BERT variant, TinyBERT). The applications of temperature rescaling and teacher model stacking further boost model accuracy, without increasing the student model complexity. We present experimental results on both in-house e-commerce search relevance data as well as a public data set on sentiment analysis from the GLUE benchmark. The latter takes advantage of another related public data set of much larger scale, while disregarding its potentially noisy labels. Embedding analysis and case study on the in-house data further highlight the strength of the resulting model. By making the data processing and model training source code public, we hope the techniques presented here can help reduce energy consumption of the state of the art Transformer models and also level the playing field for small organizations lacking access to cutting edge machine learning hardwares.

This paper proposes a new HDP based online review rating regression model named Topic-Sentiment-Preference Regression Analysis (TSPRA). TSPRA combines topics (i.e. product aspects), word sentiment and user preference as regression factors, and is able to perform topic clustering, review rating prediction, sentiment analysis and what we invent as "critical aspect" analysis altogether in one framework. TSPRA extends sentiment approaches by integrating the key concept "user preference" in collaborative filtering (CF) models into consideration, while it is distinct from current CF models by decoupling "user preference" and "sentiment" as independent factors. Our experiments conducted on 22 Amazon datasets show overwhelming better performance in rating predication against a state-of-art model FLAME (2015) in terms of error, Pearson's Correlation and number of inverted pairs. For sentiment analysis, we compare the derived word sentiments against a public sentiment resource SenticNet3 and our sentiment estimations clearly make more sense in the context of online reviews. Last, as a result of the de-correlation of "user preference" from "sentiment", TSPRA is able to evaluate a new concept "critical aspects", defined as the product aspects seriously concerned by users but negatively commented in reviews. Improvement to such "critical aspects" could be most effective to enhance user experience.