Recommender systems (RS) are currently being studied to mitigate limitations during cold-start conditions by leveraging modality information or introducing Agent concepts based on the exceptional reasoning capabilities of Large Language Models (LLMs). Meanwhile, food and beverage recommender systems have traditionally used knowledge graph and ontology concepts due to the domain's unique data attributes and relationship characteristics. On this background, we propose MARC, a multimodal and multi-task cocktail recommender system based on Agentic Retrieval-Augmented Generation (RAG) utilizing graph database under cold-start conditions. The proposed system generates high-quality, contextually appropriate answers through two core processes: a task recognition router and a reflection process. The graph database was constructed by processing cocktail data from Kaggle, and its effectiveness was evaluated using 200 manually crafted questions. The evaluation used both LLM-as-a-judge and human evaluation to demonstrate that answers generated via the graph database outperformed those from a simple vector database in terms of quality. The code is available at https://github.com/diddbwls/cocktail_rec_agentrag

Concurrent to the rapid progress in the development of neural-network based models in areas like natural language processing and computer vision, the need for creating explanations for the predictions of these black-box models has risen steadily. We propose a new method to generate extractive explanations for predictions made by neural networks, that is based on masking parts of the input which the model does not consider to be indicative of the respective class. The masking is done using gradient-based optimization combined with a new regularization scheme that enforces sufficiency, comprehensiveness and compactness of the generated explanation, three properties that are known to be desirable from the related field of rationale extraction in natural language processing. In this way, we bridge the gap between model interpretability and rationale extraction, thereby proving that the latter of which can be performed without training a specialized model, only on the basis of a trained classifier. We further apply the same method to image inputs and obtain high quality explanations for image classifications, which indicates that the conditions proposed for rationale extraction in natural language processing are more broadly applicable to different input types.

Generating safe and non-conservative behaviors in dense, dynamic environments remains challenging for automated vehicles due to the stochastic nature of traffic participants' behaviors and their implicit interaction with the ego vehicle. This paper presents a novel planning framework, Multipolicy And Risk-aware Contingency planning (MARC), that systematically addresses these challenges by enhancing the multipolicy-based pipelines from both behavior and motion planning aspects. Specifically, MARC realizes a critical scenario set that reflects multiple possible futures conditioned on each semantic-level ego policy. Then, the generated policy-conditioned scenarios are further formulated into a tree-structured representation with a dynamic branchpoint based on the scene-level divergence. Moreover, to generate diverse driving maneuvers, we introduce risk-aware contingency planning, a bi-level optimization algorithm that simultaneously considers multiple future scenarios and user-defined risk tolerance levels. Owing to the more unified combination of behavior and motion planning layers, our framework achieves efficient decision-making and human-like driving maneuvers. Comprehensive experimental results demonstrate superior performance to other strong baselines in various environments.

The long-tailed class distribution in visual recognition tasks poses great challenges for neural networks on how to handle the biased predictions between head and tail classes, i.e., the model tends to classify tail classes as head classes. While existing research focused on data resampling and loss function engineering, in this paper, we take a different perspective: the classification margins. We study the relationship between the margins and logits (classification scores) and empirically observe the biased margins and the biased logits are positively correlated. We propose MARC, a simple yet effective MARgin Calibration function to dynamically calibrate the biased margins for unbiased logits. We validate MARC through extensive experiments on common long-tailed benchmarks including CIFAR-LT, ImageNet-LT, Places-LT, and iNaturalist-LT. Experimental results demonstrate that our MARC achieves favorable results on these benchmarks. In addition, MARC is extremely easy to implement with just three lines of code. We hope this simple method will motivate people to rethink the biased margins and biased logits in long-tailed visual recognition.

Recently, deep learning approaches for MR motion artifact correction have been extensively studied. Although these approaches have shown high performance and reduced computational complexity compared to classical methods, most of them require supervised training using paired artifact-free and artifact-corrupted images, which may prohibit its use in many important clinical applications. For example, transient severe motion (TSM) due to acute transient dyspnea in Gd-EOB-DTPA-enhanced MR is difficult to control and model for paired data generation. To address this issue, here we propose a novel unsupervised deep learning scheme through outlier-rejecting bootstrap subsampling and aggregation. This is inspired by the observation that motions usually cause sparse k-space outliers in the phase encoding direction, so k-space subsampling along the phase encoding direction can remove some outliers and the aggregation step can further improve the results from the reconstruction network. Our method does not require any paired data because the training step only requires artifact-free images. Furthermore, to address the smoothing from potential bias to the artifact-free images, the network is trained in an unsupervised manner using optimal transport driven cycleGAN. We verify that our method can be applied for artifact correction from simulated motion as well as real motion from TSM successfully, outperforming existing state-of-the-art deep learning methods.

Following the amazing turn in of redditors for previous lectures, we are organizing another free zoom lecture for the reddit community. In this next lecture Dr. Marc Eder will talk about his research - Maximizing Computer Visions's Field. This talk will introduce the emerging field of 360 computer vision, and provide an overview of the spherical distortion problem, highlighting how this distortion affects many of the highest profile problems in computer vision, from deep learning to structure-from-motion and SLAM. It will survey some of the existing work on the topic, and identify 3 guiding principles that drive a general solution to the problem. Finally, we will conclude with some opportunities for further research and some big picture takeaways from work thus far.

Marc Carrel-Billiard is the Global Senior Managing Director of Accenture Labs, the company's dedicated R&D organization. In his role, he also directs Accenture's annual Technology Vision research, which looks at the future of enterprise technology. Marc has been with Accenture for nearly 20 years and has worked across all the five industries we serve. Before taking on leadership of Technology R&D, Marc was the global lead for Emerging Technology in Accenture. He has held several global leadership roles within Accenture's technology group, including within Application Portfolio Optimization and SOA/Integration Architecture.

Both artificial intelligence and machine learning are trending for 2019 with absolutely no signs of slowing down anytime soon. While this new era of technology has a tremendous number of benefits, it can sometimes be difficult to tell fact from fiction. Recently, our Co-founder Marc Poirier, along with PPC pros Brad Geddes of AdAlysis and Jeff Allen of Hanapin Marketing hosted a webinar to address these questions. During the webinar they covered many questions on AI and machine learning, and we're sharing the hottest questions in the round up below. Brad shared how over the past five years, the two biggest things in paid search have been machine learning -- which is all about automation -- and then audience targeting where we want to get specific about each user group.

In this special guest feature, Marc Alacqua, CEO and founding partner of Signafire, discusses a useful approach to data – known as data fusion – which is essentially alchemy-squared, turning not just one but multiple raw materials in to something greater than the sum of their parts. It goes beyond older methods of big data analysis, like data integration, in which large data sets are simply thrown together in one environment. Marc is a decorated combat veteran of the U.S. Army Special Operations Forces. For his service during Operation Iraqi Freedom, he was cited for "exceptionally conspicuous gallantry" and awarded two Bronze Star Medals and the Army Commendation Medal for Valor. A 20-year veteran and Lieutenant Colonel, Marc has extensive command experience in both combat and peace time, having commanded airborne and light infantry as well as special operations units.

It seems there are few jobs robots can't do these days, even the most delicate jobs, like picking asparagus or potting plant seedlings. But they're only needed because humans can't - or won't - do the work, farmers say. Marc Vermeer had a problem. He was struggling to attract workers to pick his white asparagus crop in the Netherlands. The workers he did hire moved on quickly, so he was always training new people.