Objective 2.3 Work with Textual, Numerical, Audio, or Video Data Formats AIP-210 Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A quantitative trading firm is building a model to predict stock price movements. The input features include 1,000 numerical indicators per stock per minute. The data is stored as float64 in a 2D array of shape (1,000,000 minutes, 1,000 features). The team has 16 GB of RAM. The model training requires the data to be in float32. What is the most appropriate approach?

   • AConvert the float64 data to float16 to save even more memory
   • BKeep the data as float64 and train the model with reduced batch size
   • CLoad the float64 data and cast it to float32 in memory
   • DLoad the data in chunks, cast each chunk to float32, and save to a memory-mapped float32 array
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A manufacturing company collects vibration data from industrial machines using high-frequency accelerometers. The raw data is stored as 24-bit integer samples at 50 kHz. The ML team wants to train a model to predict equipment failure. They plan to compute frequency-domain features. What is the most appropriate first step?

   • AConvert the data to 8-bit integers to reduce memory
   • BApply the FFT directly to the 24-bit integer samples
   • CConvert the 24-bit integers to float32 and apply a Fast Fourier Transform (FFT)
   • DDownsample the data to 10 kHz before any processing
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A climate-research institute stores daily temperature readings from 5,000 weather stations as a matrix where rows are stations and columns are days. The matrix is 5,000 x 10,000 (50 million float64 values). The team needs to train a model that accesses entire rows (all days for one station) frequently. Which storage layout is most appropriate?

   • AStore the matrix as a JSON array of arrays
   • BRow-major (C-order) layout
   • CColumn-major (Fortran-order) layout
   • DStore the matrix as a list of Python lists
   Reveal Solution Discussion Select an answer first
5. 19foundation · easy

   Which numerical data structure is most appropriate for storing a two-dimensional grid of numbers, such as a grayscale image where each pixel is a value?

   • AFloat
   • BInteger
   • CMatrix
   • DArray
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   Which conversion is commonly performed to transform raw audio into a visual representation that can be used as input to a convolutional neural network for speech recognition?

   • AAudio to spectrogram
   • BCSV to JSON
   • CText to numerical vectors
   • DVideo to frames
   Reveal Solution Discussion Select an answer first