Experience
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Education
-
-
-
Publications
-
Deep Learning for Anomaly Detection (Tutorial)
WSDM
See publicationAnomaly detection has been widely studied and used in diverse applications. Building an effective anomaly detection system requires the researchers/developers to learn the complex structure from noisy data, identify the dynamic anomaly patterns and detect anomalies while lacking sufficient labels. Recent advancement in deep learning techniques has made it possible to largely improve anomaly detection performance compared to the classical approaches. This tutorial will help the audience gain a…
Anomaly detection has been widely studied and used in diverse applications. Building an effective anomaly detection system requires the researchers/developers to learn the complex structure from noisy data, identify the dynamic anomaly patterns and detect anomalies while lacking sufficient labels. Recent advancement in deep learning techniques has made it possible to largely improve anomaly detection performance compared to the classical approaches. This tutorial will help the audience gain a comprehensive understanding of deep learning-based anomaly detection techniques in various application domains. First, it introduces what is the anomaly detection problem, the approaches taken before the deep model era and the challenges it faced. Then it surveys the state-of-the-art deep learning models extensively and discusses the techniques used to overcome the limitations from traditional algorithms. Second to last, it studies deep model anomaly detection techniques in real world examples from LinkedIn production systems. The tutorial concludes with a discussion of future trends.
-
AI Model Monitoring
2019 Grace Hopper Celebration
In order to detect issues in AI models and minimize negative impact on LinkedIn business metrics, an end-to-end monitoring system for machine learning models is needed. Within this document, we define a set of metrics to quantify AI models’ issues, and demonstrate technical details for detecting and summarizing signals for model performance degradation and feature distribution changes.
Patents
-
Machine Learning Model Monitoring
Issued US 902408-US-NP
-
Auto-Tune Anomaly Detection
Issued US 902269-US-NP
Auto-tuning is using a logistic classifier model to trade-off between severity and duration to classify whether a detected anomaly is a true anomaly or not. The model is capable to incorporate users’ feedbacks on historical anomalies, select cutoff based on best precision given maximal recall. Auto-tune is also optimized based on user-customized constraints on "minimum-time-to-alert" anomalies.
Other similar profiles
Explore more posts
-
Jessica H.
OSS Bench for Agentic Coding.. in steps Cline-Bench: "A task that frontier LLMs struggle to complete are ideal candidates for cline-bench. If you find yourself having to intervene, or write code manually because cline was not able to complete the job - your task captures the failure boundary of state-of-the-art models and is precisely what cline-bench seeks to formalize." https://lnkd.in/gHcGYxiZ
-
Vikas Chandra
𝐒𝐮𝐛-𝐁𝐢𝐥𝐥𝐢𝐨𝐧 𝐑𝐞𝐚𝐬𝐨𝐧𝐢𝐧𝐠 𝐃𝐢𝐝𝐧'𝐭 𝐒𝐭𝐚𝐫𝐭 𝐰𝐢𝐭𝐡 𝐑𝐋 MobileLLM-R1, our 950M-parameter model, matches Qwen3-0.6B on MATH500 (74.0 vs 73.0) while using 11.7% of its pretraining data. That result didn't come from RL alone. It sits on a stack we've been building since 2023: architecture design (MobileLLM), quantization (SpinQuant, LLM-QAT, ParetoQ), data curation (Target-Aware LM, AutoMixer), and finally RL post-training (MobileLLM-R1). Remove any layer and the result breaks. I wrote up how these pieces fit together: why deeper-and-thinner beats balanced scaling below 1B, how influence-driven data mixing lets you match 36T-token models with 4.2T tokens, and why naive RL doesn't work at sub-billion scale. Also where we're extending this to video (VideoAuto-R1), 3D vision (DepthLM), and audio (EgoAVU, SLAP). Full post: https://lnkd.in/gZQmVT-k
1 Comment
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content