MLconf SF 2019
2019-11-08
San Francisco, United States
[MLconf SF 2014, MLconf SF 2015, MLconf SF 2016, MLconf SF 2017, MLconf SF 2019]
who it's for
machine learning engineers, data scientists, applied researchers, natural language processing researchers, computer vision engineers, healthcare ML engineers, data engineers, machine learning product managers
topics
automated forecasting and anomaly platforms; in-dumpster camera CNN pipelines; contamination scoring metrics; weak supervision with Snorkel; dataset versioning and labeling quality; large-scale IMU gait datasets; prototypical clustering networks for few-shot diagnosis; temporal convolutional networks and Transformer-XL alignment; TensorFlow 1.x real-time model IDE; modeling observation probability p(obs|Y)
| title | author | length |
|---|---|---|
| Uber’s Intelligent Insights Assistant | Franziska Bell | 26m |
Uber reports platforms that turn internal users into analysts: a forecasting platform that auto-preprocesses, backtests and selects models (classical, ML… | ||
| Applying Computer Vision to Reduce Contamination in the Recycling Stream | Justin Armstrong | 13m |
Compology deploys in-dumpster cameras that upload images to a cloud pipeline using a convolutional neural network plus human review to count and time-stamp… | ||
| Understanding Human Impact: Social and Equity Assessments for AI Technologies | Jamila Smith-Loud | 21m |
Describes a concrete framework that adapts environmental impact assessments into social and equity impact assessments for ML products, backed by hiring social… | ||
| Quantum Computing: a Treasure Hunt, not a Gold Rush | Igor Markov | 24m |
Quantum computers currently show narrow, special-case advantages: Google claims 53 qubits (54 fabricated, 1 failed) for a sampling task, but that task is an… | ||
| Data Labeling as Religious Experience | Josh Wills | 25m |
Describes building a spam classifier for Slack invites that fails when using invite-acceptance as the label and reveals the comical result that VP of… | ||
| Project GaitNet: Ushering in the ImageNet moment for human Gait kinematics | Vinay Prabhu | 24m |
A planet-sized IMU dataset of roughly 1.2 million gait cycles from volunteers across 170 countries is segmented by heel strikes into N×4×100 tensors… | ||
| AI for healthcare: Scaling Access and Quality of Care for Everyone | Anitha Kannan, Xavier Amatriain | 25m |
Curai deploys a mobile primary-care app with hundreds of thousands of users that combines NLP, vision and automatic diagnosis by mixing legacy expert systems… | ||
| The Brain’s Guide to Dealing with Context in Language Understanding | Ted Willke | 43m |
Argues that language regions form a hierarchy of timescales and shows that modern models mirror this: compares LSTMs, temporal convolutional network (TCN) and… | ||
| How to Enable Warp Speed for Machine Learning Modeling | Martin Isaksson | 15m |
A drag-and-drop IDE that builds, trains and debugs models while exposing the underlying code and runtime variables, currently parsing and running TensorFlow… | ||
| The Importance of Modeling Data Collection | Noam Finkelstein | 18m |
Argues that systematic observation bias skews ML estimates and routine validation hides it, shows two concrete fixes: model the observation probability as… | ||
| The Uncanny Valley of ML | June Andrews | 25m |
Argument: adding ML recommendations to human decision workflows can lower overall accuracy and slow decisions as model accuracy approaches humans, an uncanny… | ||
| Deep Learning Architectures for Semantic Relation Detection Tasks | Sneha Rajana | 26m |
Creates a ~200k-pair labeled set by expanding WordNet (3k direct → +14k indirect) and mining PPDB paraphrases (80k from not‑XY → +25k → +81k after expansion)… | ||
| ‘The Voice’: New Challenges in a Zero UI world | Vito Ostuni | 19m |
Pandora launches an in-app voice mode serving part of a 100+ million monthly user base and classifies queries as known-item, thematic, or broad open-ended. It… | ||
| Building an Incrementally Trained, Local Taste Aware, Global Deep Learned Recommender System Model | Anoop Deoras | 18m |
Netflix scale (about 150M accounts) drives a global recommender built as a variational autoencoder that encodes country, device and time by concatenating… | ||
| Optimized Image Classification on the Cheap | Meghana Ravikumar | 17m |
Stanford Cars (≈16k images, 196 classes) compares two transfer methods: ResNet18 fine-tune versus feature extraction from ResNet50. Hyperparameter search uses… | ||
| Machine Learning Methods in Detecting Alzheimer’s Disease from Speech and Language | Jekaterina Novikova | 27m |
Winterlight Labs builds a classification pipeline that combines acoustic features (pitch energy pauses MFCCs) with lexical and syntactic features from… | ||
| nothing matches | ||