Neural Network Implementations

Problem

Different classification and time-series datasets offer a practical way to investigate neural network architectures.

Objective

Implement and experiment with Perceptron, MLP, CNN, and RNN architectures.

Approach

Apply neural network techniques to Iris, Breast Cancer, handwritten-digit, and Yahoo Finance datasets.

Implementation

Implemented and experimented with four architectures across classification and financial time-series datasets, as described in the AI-focused CV.

Challenges to investigate

Architecture and input representation should fit the task. The source does not document training configurations or specific debugging challenges.

The source does not document the project’s actual obstacles; the following is a technical consideration, not a retrospective claim.

Documented outcomes

Four architecture implementations are documented. No comparative accuracy, financial returns, or production deployments are claimed.

Technical reflection

Working across different input structures helps clarify the relationship between a learning task and the model architecture.

This reflection follows from the documented method; it is not presented as a measured project outcome.