From the first row to the final result. Modules included with Aner share native types, explicit contracts, and a common language.
These modules currently ship with Aner. Independent packages and distribution of libraries from other developers are part of the roadmap.
01Built in
aner.nn
Describe the model. See the learning.
Build small dense neural networks in Aner with a concise sequential API or an explicit training loop. The first backend uses native CPU Float64 tensors.
Dense layers and tanh activation
Automatic differentiation in reverse mode
SGD and softmax cross entropy
Seeded initialization and fitted classifier snapshots
Declare CSV schemas, preserve identifiers and missing values, and inspect a lazy scan before execution. Managed buffer budgets make collection an explicit decision.
Typed CSV schemas and exact missing tokens
Projection and bounded filter operations
Batch summaries and explicit collection
Validation of the full schema before filtering
Joins, Parquet, SQL, and streaming model training remain planned
An offline teaching catalogue starts with Iris and Wine, including citations, source hashes, feature metadata, deterministic stratified splits, and preprocessing fitted only on training data.
Classical and neural workflows meet in typed fitted models. Start with KNN, K means, and concise neural classifier pipelines; inspect exact limits in the reference.
Fitted KNN classifiers
Seeded K means clustering
Neural classifier pipelines with scaling fitted only on training data
Explicit prediction and evaluation
SVM, decision trees, and general model interfaces remain planned
