Aleksei Zinovev

val conv1Weights: Variable = tf.variable(Shape.make(5L, 5L, NUM_CHANNELS, 32), Float::class.javaObjectType) val conv1Biases: Variable = tf.variable(Shape.make(32), Float::class.javaObjectType) Could be rewritten in the tf.variable(Shape.make(32)) via extension functions inline with reified types

1. DL4j 2. TF 3. DJL (if it is possible)

1. Tensor calculations 2. Inference (with MNIST) 3. Graph for forward pass (optional) 4. LinReg (optional) 5. MNIST (2-3 layers) 6. LSTM (optional)

Add column name repairing for all IO sources as a common step for dataframe creation

2

The solution could be based on the PR #386

Prepare a survey (or GitHub Discussion) about data sources

12

The draft list of data sources: 1. SQL Databases based on JDBC 2. XML 3. Protobuf 4. Parquet 5. ORC 6. SparkSQL 7. different files on the FileSystem 8. NoSQL...

- [ ] PGbox - [ ] PGcircle - [ ] PGpath

Make examples for database runnable as a part of TC

1

Now, the example project `examples-jdbc` is commented in `settings.gradle.kts`.

From @Jolanrensen > DataFrame.readResultSet(resultSet, PostgreSql) > cannot be called more than once! This is not expected behavior for me as dataframe calls are always functional. They do not modify the...

- [ ] common link with in the GET with parameters style is not supported correctly (add a test for this) - [ ] common url with login and password