Learn Before
Sufficient Data and Model Capacity Can Remove Feature Bottlenecks
If an end-to-end model is large enough and trained on enough labeled examples, it can learn useful patterns directly from the raw input. In that situation, it is less constrained by hand-built intermediate representations such as MFCCs or phoneme labels, and its error can move closer to the best achievable level for the task.
0
1
Tags
Python Programming Language
Data Science
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Direct Sentiment Prediction from Raw Reviews
End-to-End Speech Recognition
Choosing a Modular Driving System
End-to-End Control Requires Many Labeled Examples
Sufficient Data and Model Capacity Can Remove Feature Bottlenecks
Learning Structured Outputs End to End
What Does End-to-End Learning Usually Replace?
Neural networks are often used in end-to-end learning pipelines.
End-to-end learning connects the _____ to the target output.
Match each model output with the kind of result it describes.
End-to-End Text Classification Flow
With the right labeled examples, what can an end-to-end model sometimes learn to produce?
End-to-end deep learning can only produce a single numerical output.
End-to-end models can learn structured outputs directly.
Match each learning setup to its definition.
Order the reasoning steps showing how end-to-end learning can produce structured outputs.
How Labeled Pairs Shape the Outputs an End-to-End Model Can Learn
Training an Image-to-Caption Model
Complex Outputs from End-to-End Models
Learn After
Scaling Up to Remove Hand-Crafted Features
Large End-to-End Models and Error Limits
Requirements for End-to-End Model Performance
Match each item to its role in an end-to-end speech system.
From Handcrafted Features to End-to-End Learning
How Model Size and Data Affect Feature Engineering
Moving to an End-to-End Speech Model
Requirements for an End-to-End Model
When End-to-End Learning Reduces Manual Feature Limits
Can a Small Network Always Stay Low-Bias with More Data?