Learn Before
Close the distribution gap in a wake-word detector
Case context: You are building a wake-word detector for a voice assistant. Your training set contains clean recordings made in a quiet studio using a desktop microphone. Your development set comes from household devices placed in kitchens, where recordings include echo, running water, and occasional clatter. The model performs well on training data but poorly on the development set.
Question: What data modification should you make to improve development-set performance, and why?
Sample answer: Add simulated kitchen echo and background noise to the clean training recordings. This makes the training distribution closer to the development distribution, so the model learns features that are more robust to the conditions seen on the development set.
Key points:
- Add simulated echo and household noise to the clean training audio.
- Make the training data resemble the development data.
- Reduce the gap caused by the noisy kitchen recording environment.
Rubric: Learners should identify adding simulated echo and background noise to the training set and explain that it narrows the distribution gap between the clean training audio and the noisy development audio.
0
1
Tags
Machine Learning
Deep Learning
Supervised Learning
Dive into Deep Learning @ D2L
Data Science
Machine Learning Strategy
Machine Learning Yearning @ DeepLearning.AI
Related
Why are validation images blurrier than training images in a parcel-scanner project?
True or False: Adding synthetic motion blur to clear training images can help the training data resemble a blurry dev set more closely.
To make training photos more similar to a dev set with dim lighting, you can add simulated _____ to clean training images.
Why Add Synthetic Blur?
Motion blur in a flower-photo validation set can come from people taking pictures while walking or holding the camera unsteadily.
To narrow the gap, add _____ to the clean training photos.
Match each data source or technique to its role in a fruit defect detector example.
Order the steps for adding synthetic blur to help a photo classifier handle blurry validation images.
What mismatch is addressed by adding synthetic blur to the training set in this classifier example?
Web-sourced training photos usually contain the same amount of handheld blur as smartphone photos collected for the development set.
The clean training images used before synthetic blur are drawn from _____ images.
Match each observation in a smart-speaker speech recognizer scenario to the corresponding explanation or action.
Order the steps in deciding whether to add synthetic blur to factory inspection photos.
Explain how synthetic blur can reduce a training–validation mismatch
Close the distribution gap in a wake-word detector
Why add synthetic rain to clear training images?