Code for IoT Journal paper 'ML-MCU: A Framework to Train ML Classifiers on MCU-based IoT Edge Devices'

PyTorch/Tensorflow solutions for Stanford's CS231n: "CNNs for Visual Recognition"

Implementation of (overlap) local SGD in Pytorch

Computer Vision and Image Processing algorithms implemented using OpenCV, NumPy and MatPlotLib, for UOM's EN2550 Fundamentals of Image Processing and Machine Vision Module ❄

A compressed adaptive optimizer for training large-scale deep learning models using PyTorch

Lookahead optimizer ("Lookahead Optimizer: k steps forward, 1 step back") for tensorflow

📈Implementing the ADAM optimizer from the ground up with PyTorch and comparing its performance on six 3-D objective functions (each progressively more difficult to optimize) against SGD, AdaGrad, and RMSProp.

Communication-efficient decentralized SGD (Pytorch)

Implement a Neural Network trained with back propagation in Python

Implement some ML algorithms in scala

Simple MATLAB toolbox for deep learning network: Version 1.0.3

JAX compilation of RDDL description files, and a differentiable planner in JAX.

MetaPerceptron: A Standardized Framework For Metaheuristic-Driven Multi-layer Perceptron Optimization

Nadir: Cutting-edge PyTorch optimizers for simplicity & composability! 🔥🚀💻

基于粒子群PSO+随机梯度下降SGD优化器的Pytorch训练框架

ND-Adam is a tailored version of Adam for training DNNs.

基于 UNSW-NB15 的双阶段贝叶斯融合入侵检测系统。从零手写 梯度下降 (GD)、SGD、Adam 等基础优化算法，并独立实现 逻辑回归、高斯朴素贝叶斯、决策树、多层感知机 (MLP) —— 无任何第三方深度学习框架。 系统采用 两阶段级联架构：Stage 1 二分类判定攻击/正常，Stage 2 九分类细分攻击类型，最终通过 贝叶斯后验概率融合 输出 10 类检测结果。在预处理中引入 数值清洗、偏度变换、类别降基及交互特征工程，并结合 分层重采样 缓解严重类别不平衡。训练支持早停与 L2 正则化。端到端评估涵盖 ROC/PR 曲线、10×10 混淆矩阵、Top‑K 置信度分析及误差案例诊断，最终 10 类 macro‑F1 达 0.44，所有优化与模型代码完全手写！

𝗔𝗡𝗡 𝗮𝗻𝗱 𝗖𝗡𝗡 𝗳𝗿𝗼𝗺 𝘀𝗰𝗿𝗮𝘁𝗰𝗵 | 𝗡𝗼 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀, 𝗷𝘂𝘀𝘁 𝗽𝘂𝗿𝗲 𝗡𝘂𝗺𝗽𝘆

Object recognition AI using deep learning

Tensorflow-Keras callback implementing arXiv 1712.07628