{"id":31642282,"url":"https://github.com/mihaibudurean/neuralnetworkscomparativeanalysis","last_synced_at":"2026-04-18T07:33:15.257Z","repository":{"id":314112656,"uuid":"1050467633","full_name":"MihaiBudurean/NeuralNetworksComparativeAnalysis","owner":"MihaiBudurean","description":"Comparative study of Network-in-Network (NiN) and GoogLeNet-style multi-branch networks on Sign-MNIST and MNIST transfer learning.","archived":false,"fork":false,"pushed_at":"2025-09-10T14:32:51.000Z","size":750,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-09-10T18:37:46.214Z","etag":null,"topics":["cnn","deep-learning","keras","multiclass-classification","tensorflow"],"latest_commit_sha":null,"homepage":"","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/MihaiBudurean.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-09-04T13:25:43.000Z","updated_at":"2025-09-10T14:32:55.000Z","dependencies_parsed_at":"2025-09-10T18:37:48.880Z","dependency_job_id":"9cce4fe0-6cf3-4257-9346-e4a81b725e3b","html_url":"https://github.com/MihaiBudurean/NeuralNetworksComparativeAnalysis","commit_stats":null,"previous_names":["mihaibudurean/neuralnetworkscomparativeanalysis"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/MihaiBudurean/NeuralNetworksComparativeAnalysis","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MihaiBudurean%2FNeuralNetworksComparativeAnalysis","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MihaiBudurean%2FNeuralNetworksComparativeAnalysis/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MihaiBudurean%2FNeuralNetworksComparativeAnalysis/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MihaiBudurean%2FNeuralNetworksComparativeAnalysis/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/MihaiBudurean","download_url":"https://codeload.github.com/MihaiBudurean/NeuralNetworksComparativeAnalysis/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/MihaiBudurean%2FNeuralNetworksComparativeAnalysis/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":278717449,"owners_count":26033542,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-10-07T02:00:06.786Z","response_time":59,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["cnn","deep-learning","keras","multiclass-classification","tensorflow"],"created_at":"2025-10-07T03:57:52.506Z","updated_at":"2025-10-07T03:58:00.599Z","avatar_url":"https://github.com/MihaiBudurean.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Comparative Analysis of NiN and Multi‑Branch Networks\n\nThis project presents a comparative study of **Network‑in‑Network (NiN)** and a **multi‑branch GoogLeNet‑style architecture** for multiclass image classification and transfer learning. It was developed as a case study using the **Sign‑MNIST** dataset and extended to **MNIST** for transfer learning.\n\n---\n\n## 📊 Project Description\n\nThe goal is to evaluate two different deep neural architectures in terms of accuracy, convergence, parameter efficiency, and transferability.\n\n### Datasets\n\n* **Sign‑MNIST**\n* **MNIST**\n\n### Architectures\n\n* **Network‑in‑Network (NiN)** – Uses 1×1 convolutions (mini‑MLPs) within convolutional layers to enhance nonlinear feature learning.\n* **Multi‑Branch Network (GoogLeNet‑style)** – Employs parallel convolutional branches (1×1, 3×3, 5×5, pooling) to capture multi‑scale features.\n\n### Tasks\n\n1. Train and evaluate NiN and GoogLeNet on **Sign‑MNIST**.\n2. Apply **transfer learning** by adapting both models to **MNIST**.\n3. Compare performance using accuracy, precision, recall, F1‑score, and resource requirements.\n\n---\n\n## 🛠 Requirements\n\nInstall the dependencies to run the project:\n\n```bash\npip install -r requirements.txt\n```\n\n---\n\n## 📈 Results\n\n* On **Sign‑MNIST**:\n\n  * NiN achieved ≈ **98.37% accuracy** with \\~180k parameters.\n  * GoogLeNet achieved ≈ **93.61% accuracy** with \\~37k parameters.\n* On **MNIST (Transfer Learning)**:\n\n  * NiN achieved ≈ **82.88% accuracy**.\n  * GoogLeNet achieved ≈ **92.64% accuracy**.\n\n**Key Insight:** NiN excels on the original dataset with higher accuracy, but GoogLeNet generalizes better in transfer learning with fewer parameters.\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmihaibudurean%2Fneuralnetworkscomparativeanalysis","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmihaibudurean%2Fneuralnetworkscomparativeanalysis","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmihaibudurean%2Fneuralnetworkscomparativeanalysis/lists"}