Exploring ML and Embedded Systems (Day 2 - Test 1)

What is TensorFlow Lite used for?

• TensorFlow Lite is used for deploying machine learning models on mobile and embedded devices.

• TensorFlow Lite is used for training large-scale models on cloud servers.

• TensorFlow Lite is used for data visualization and analysis.

• TensorFlow Lite is used for creating web applications.

Describe the process of converting a model to TensorFlow Lite.

• Using only Keras to train the model

• Exporting the model directly to ONNX format

• Running the model on a CPU without conversion

• The process involves training a model, using the TensorFlow Lite Converter, optimizing, and saving as a .tflite file.

What are the key benefits of using TensorFlow Lite for embedded systems?

• Key benefits of using TensorFlow Lite for embedded systems include reduced model size, lower latency, efficient resource usage, support for quantization, and on-device machine learning capabilities.

• Increased model complexity

• Higher power consumption

• Limited support for mobile devices

What is TinyML and how does it relate to the IoT market?

• TinyML is a type of cloud computing for large data centers.

• TinyML is a technology that enables machine learning on small devices, enhancing the capabilities of IoT devices.

• TinyML is a programming language for IoT devices.

• TinyML is a hardware component that replaces traditional sensors.

What are the primary phases in machine learning model development?

• Data Analysis, Feature Engineering, Model Testing

• Data Visualization, Hyperparameter Tuning, Model Interpretation

• Data Annotation, Model Optimization, Result Presentation

• Data Collection, Data Preprocessing, Model Selection, Model Training, Model Evaluation, Model Deployment, Monitoring and Maintenance

Explain the importance of data preprocessing in machine learning.

• Data preprocessing is only necessary for supervised learning.

• Data preprocessing is essential for improving data quality and model performance in machine learning.

• Data preprocessing has no impact on model accuracy.

• Data preprocessing is primarily used for data visualization.

What role does model evaluation play in the development process?

• Model evaluation focuses solely on the model's training data.

• Model evaluation is a one-time process that doesn't require updates.

• Model evaluation is only necessary for academic purposes.

• Model evaluation ensures the model's performance is validated and optimized for real-world application.

How can transfer learning be utilized in TinyML applications?

• Transfer learning is not applicable in TinyML due to hardware limitations.

• Transfer learning allows for efficient model adaptation in TinyML by using pre-trained models to improve performance on specific tasks with limited data.

• Transfer learning can only be used with large datasets in TinyML applications.

• TinyML requires training models from scratch for every new task.

What challenges are faced in deploying machine learning models on embedded systems?

• Challenges include limited resources, memory constraints, power consumption, real-time processing, model optimization, hardware compatibility, and deployment management.

• No need for model optimization

• High power consumption is beneficial

• Unlimited resources for processing

How does overfitting affect machine learning models?

• Overfitting has no impact on model accuracy.

• Overfitting improves model performance on all datasets.

• Overfitting leads to models that perform well on training data but poorly on unseen data.

• Overfitting is beneficial for real-time applications.