mingrad is a lean, but mean autograd engine that extends micrograd into a mini deep learning framework.

• Vectorized Operations: Execution of advanced mathematical tasks through efficient array-based processing.
• NumPy Integration: NumPy's C backend for enhanced numerical computation efficiency.
• Scalability: Batch processing and management of high-dimensional data arrays.
• NN architecture into three key operation types:
  • Unary Operations: Single-array manipulations using element-wise operations such as RELU.
  • Binary Operations: Combining two arrays to produce a single output, operations like add and mul.
  • Reduction Operations: Streamlining data by reducing larger arrays into a summarized form through operations such as mean and argmax.
• Modularity: Straightforward expansions and customizations.

• linear neural net prediction on diabetes dataset 👇🏻