Neural Network Neural Network Applications and Frameworks 2 β Questions and Answers
Question 1: What is model quantization in the context of deploying neural networks?
- Measuring model performance with specific metrics
- Reducing numerical precision of weights (e.g., from FP32 to INT8) to decrease model size and inference latency (Correct answer)
- Counting the number of parameters in a model
- Converting a model to run on quantum computing hardware
Correct answer: Reducing numerical precision of weights (e.g., from FP32 to INT8) to decrease model size and inference latency
Quantization reduces weight and activation precision from 32-bit float to lower bits (e.g., INT8), dramatically reducing model size and speeding up inference with minimal accuracy loss.
Question 2: What is neural architecture search (NAS)?
- Manually designing optimal neural network architectures
- Automated methods to discover optimal architecture configurations using search algorithms (Correct answer)
- Searching a database of pretrained architectures
- Pruning unnecessary layers from existing architectures
Correct answer: Automated methods to discover optimal architecture configurations using search algorithms
NAS uses search strategies (evolutionary, reinforcement learning, gradient-based) to automatically discover neural architecture designs that optimize performance on a target task.
Question 3: What is the purpose of mixed precision training?
- Using different activation functions for different layers
- Using FP16 for most computations while keeping FP32 for critical operations, speeding training while maintaining accuracy (Correct answer)
- Training the model with multiple different loss functions
- Applying different learning rates to different layers
Correct answer: Using FP16 for most computations while keeping FP32 for critical operations, speeding training while maintaining accuracy
Mixed precision training uses 16-bit floats (FP16) for forward/backward passes and 32-bit floats (FP32) for weight updates, achieving ~2-3x speedup with minimal accuracy loss on modern GPUs.
Question 4: What is federated learning in the context of neural network training?
- Distributing training across multiple GPUs in a data center
- Training models across decentralized devices while keeping data local and only sharing model updates (Correct answer)
- A method for federating pretrained models from multiple sources
- Distributed hyperparameter tuning across cloud providers
Correct answer: Training models across decentralized devices while keeping data local and only sharing model updates
Federated learning trains a shared model across many devices (e.g., smartphones) without centralizing data β each device trains locally and only shares weight updates with a central server.
Question 5: What is model pruning in neural network deployment?
- Removing entire hidden layers from the model
- Removing weights or neurons that contribute little to predictions, reducing model size (Correct answer)
- Deleting misclassified training examples
- Reducing the vocabulary size of NLP models
Correct answer: Removing weights or neurons that contribute little to predictions, reducing model size
Model pruning identifies and removes parameters (weights, neurons, filters) that have minimal impact on model output, creating sparser, smaller models with faster inference.
Question 6: What is the primary use case for TensorFlow Lite?
- Training large models on cloud TPUs
- Running pretrained neural network models on mobile and embedded devices with low latency (Correct answer)
- Visualizing TensorFlow model architectures
- Distributed training across multiple machines
Correct answer: Running pretrained neural network models on mobile and embedded devices with low latency
TensorFlow Lite is a lightweight runtime for deploying trained TensorFlow models on mobile (Android/iOS) and embedded (microcontrollers) devices with optimized performance.
Neural Network Neural Network Applications and Frameworks 2
What is model quantization in the context of deploying neural networks?