<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/4e3c9173-7c25-4599-bb20-eb01eaf7f192/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/4e3c9173-7c25-4599-bb20-eb01eaf7f192/0_JAXON_Logo_Mark.jpg" width="40px" /> For a short description of each field found in the Classical Model Intake Form, See the Neural Model Creation: Intake Form Selection Breakdown below.
</aside>
Model creation is asynchronous - once the job is submitted, users may close the browser window and return later to check the job status. Users may submit multiple such jobs, and each job will run in the order in which they are submitted. Neural model creation may take a few minutes, or up to several hours.
<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/915594a2-addd-4061-b9f5-42a3efb8a596/JAXON_Logo_Mark_on_blue.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/915594a2-addd-4061-b9f5-42a3efb8a596/JAXON_Logo_Mark_on_blue.jpg" width="40px" /> At least one Training Stage must be added in order to be able to select the Schedule Training button.
</aside>
Deep neural networks are trained over several epochs. During each epoch, the model processes each example in the training dataset and a loss function is calculated. Typically, training context remains the same throughout all epochs: parameters, dataset, loss function, layer management, feature vocabulary, etc. When the loss has converged, training stops. Jaxon allows users to implement custom multi-stage training stages.
Here, Jaxon provides valuable information about model training and accuracy. The bar graph indicates each Training Stage completed and the F-score for each stage.
<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/ba5c13de-a8fa-4df9-a292-84bd9bdfb95b/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/ba5c13de-a8fa-4df9-a292-84bd9bdfb95b/0_JAXON_Logo_Mark.jpg" width="40px" /> If the first stage is a pretraining stage, only Training Stages 2+ will be displayed within the bar graph.
</aside>
The information to the right of the bar chart also provides a high-level overview of the training stage and is indicative of overall model performance with cumulative totals for all available Training Stages:
Name of the model
Description of the model (if any)
F-Score
Precision
Recall
Text Representation used
Tabular Representation used
Test Dataset the model was evaluated against
A table is included that contains a breakdown of the Train dataset used for each individual training stage as well as a per-stage F-Score, Precision, and Recall score.
Click on the pie icon to display a Loss Chart for that stage.
The overview provides a Confusion Matrix that maps out the predicted vs actual accuracy of the classifier for the given examples.
<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/ca09c76a-dd0a-4e96-b4ec-b179094fe6de/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/ca09c76a-dd0a-4e96-b4ec-b179094fe6de/0_JAXON_Logo_Mark.jpg" width="40px" /> Keep in mind that the confusion matrix is created for the model based on the test dataset that was provided with the training schedule - therefore the accuracy or F-Score for the model depends on the number of examples in the test set and how good the test set is.
</aside>
Hovering over one of the cells within the confusion matrix will bring up the number of examples that were expected to fall into that cell (Predicted) vs the number of examples that actually fell within that cell (Actual).
The hover box provided for the top left cell of the confusion matrix.
Clicking one of the cells within the confusion matrix will bring up actual examples that fell into that cell. This can be used for further calibration to see why certain mistakes happened and strategize how to resolve said mistakes.
The examples shown that fell into the top left cell of the confusion matrix.
Users also have the option to Import an existing external model into the Jaxon Platform. Only models created by and exported from Jaxon can be imported, which allows models to move between projects or Jaxon servers. The primary intended use case for this is to spend much time pretraining a model on lots of company data, then be able to reuse it in other projects.
Copying a neural model is similar to creating a new model. Using the same intake form, users can curate and fine-tune extant models for a given dataset.
<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/4e3c9173-7c25-4599-bb20-eb01eaf7f192/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/4e3c9173-7c25-4599-bb20-eb01eaf7f192/0_JAXON_Logo_Mark.jpg" width="40px" /> For a short description of each field found in the Neural Model Intake Form, See the Neural Model Creation: Intake Form Selection Breakdown above
</aside>
<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/294996b8-c9d0-41cc-8c27-2175066d5b88/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/294996b8-c9d0-41cc-8c27-2175066d5b88/0_JAXON_Logo_Mark.jpg" width="40px" /> You cannot change the Tabular or Text Representations from those of the original neural model.
</aside>
