Introduction to Tensors and Numpy

Overview

• Tensors are foundational data structures in deep learning, used in linear algebra.
• Similar to vectors and matrices, tensors allow for mathematical operations.
• TensorFlow, Google's machine learning library, prominently features tensors.

Tensors in TensorFlow

• Definition: In TensorFlow, a tensor is considered a multi-dimensional array.
• Conversion:
  • Tensors can be converted to and from Numpy arrays.
  • TensorFlow objects have data types and shapes similar to Numpy ndarrays.
• Acceleration:
  • Tensors can be accelerated using GPU or TPU for improved performance.

TensorFlow Operations

• TensorFlow provides rich operations libraries:
  • tf.add for addition.
  • tf.matmul for matrix multiplication.
• Native Python types are automatically converted in these operations.
• Examples:
  • Addition: Adding vectors using tf.add.
  • Multiplication: Matrix multiplication with tf.matmul.

Understanding Numpy

• Tensors are akin to Numpy arrays concerning operations.
• Axes, Rank, and Dimensions:
  • 1D array: Similar to a vector, created with one layer of brackets.
  • 2D array: Like a matrix, created with two layers of brackets.
  • 3D array: Higher dimensional arrays created by extending the bracket layers.
• Shapes:
  • Fixed structures are defined by shapes.
  • Example: A shape "two, three" indicates two subarrays, each with three elements.

Numpy Arrays

• Multi-Dimensional Arrays:
  • Created using multiple bracket layers and Numpy's np.array function.
• Dimensions and Axes:
  • Dimensions represent the number of elements in axes.
  • A 2D array with shape "two, three" has two axes and dimensions defined by the number of elements.

Terminology

• 1D Array: Vector
• 2D Array: Matrix
• 3D Array: Array with three axes

Conclusion

• Quick introduction to tensors and their relationship with Numpy.
• Next video will cover basic properties and attributes of Numpy.