Examples with Layer com.simiacryptus.mindseye.lang.Layer used on opensource projects

Example 26 with Layer

use of com.simiacryptus.mindseye.lang.Layer in project MindsEye by SimiaCryptus.

the class L2NormalizationTest method train.

@Override
public void train(@Nonnull final NotebookOutput log, @Nonnull final Layer network, @Nonnull final Tensor[][] trainingData, final TrainingMonitor monitor) {
    log.p("Training a model involves a few different components. First, our model is combined mapCoords a loss function. " + "Then we take that model and combine it mapCoords our training data to define a trainable object. " + "Finally, we use a simple iterative scheme to refine the weights of our model. " + "The final output is the last output value of the loss function when evaluating the last batch.");
    log.code(() -> {
        @Nonnull final SimpleLossNetwork supervisedNetwork = new SimpleLossNetwork(network, new EntropyLossLayer());
        @Nonnull final Trainable trainable = new L12Normalizer(new SampledArrayTrainable(trainingData, supervisedNetwork, 1000)) {

            @Override
            public Layer getLayer() {
                return inner.getLayer();
            }

            @Override
            protected double getL1(final Layer layer) {
                return 0.0;
            }

            @Override
            protected double getL2(final Layer layer) {
                return 1e4;
            }
        };
        return new IterativeTrainer(trainable).setMonitor(monitor).setTimeout(3, TimeUnit.MINUTES).setMaxIterations(500).runAndFree();
    });
}

Example 27 with Layer

use of com.simiacryptus.mindseye.lang.Layer in project MindsEye by SimiaCryptus.

the class NLayerTest method test.

/**
 * Test.
 *
 * @param log the log
 */
public void test(@Nonnull final NotebookOutput log) {
    log.h1("%s", getClass().getSimpleName());
    @Nonnull final int[] inputDims = getInputDims();
    @Nonnull final ArrayList<int[]> workingSpec = new ArrayList<>();
    for (final int[] l : dimList) {
        workingSpec.add(l);
        @Nonnull final Layer layer = buildNetwork(concat(inputDims, workingSpec));
        graphviz(log, layer);
        test(log, layer, inputDims);
    }
}

Example 28 with Layer

use of com.simiacryptus.mindseye.lang.Layer in project MindsEye by SimiaCryptus.

the class MnistTestBase method validate.

/**
 * Validate.
 *
 * @param log     the log
 * @param network the network
 */
public void validate(@Nonnull final NotebookOutput log, @Nonnull final Layer network) {
    log.h1("Validation");
    log.p("If we apply our model against the entire validation dataset, we get this accuracy:");
    log.code(() -> {
        return MNIST.validationDataStream().mapToDouble(labeledObject -> predict(network, labeledObject)[0] == parse(labeledObject.label) ? 1 : 0).average().getAsDouble() * 100;
    });
    log.p("Let's examine some incorrectly predicted results in more detail:");
    log.code(() -> {
        @Nonnull final TableOutput table = new TableOutput();
        MNIST.validationDataStream().map(labeledObject -> {
            final int actualCategory = parse(labeledObject.label);
            @Nullable final double[] predictionSignal = network.eval(labeledObject.data).getData().get(0).getData();
            final int[] predictionList = IntStream.range(0, 10).mapToObj(x -> x).sorted(Comparator.comparing(i -> -predictionSignal[i])).mapToInt(x -> x).toArray();
            // We will only examine mispredicted rows
            if (predictionList[0] == actualCategory)
                return null;
            @Nonnull final LinkedHashMap<CharSequence, Object> row = new LinkedHashMap<>();
            row.put("Image", log.image(labeledObject.data.toGrayImage(), labeledObject.label));
            row.put("Prediction", Arrays.stream(predictionList).limit(3).mapToObj(i -> String.format("%d (%.1f%%)", i, 100.0 * predictionSignal[i])).reduce((a, b) -> a + ", " + b).get());
            return row;
        }).filter(x -> null != x).limit(10).forEach(table::putRow);
        return table;
    });
}

Example 29 with Layer

use of com.simiacryptus.mindseye.lang.Layer in project MindsEye by SimiaCryptus.

the class MnistTestBase method report.

/**
 * Report.
 *
 * @param log            the log
 * @param monitoringRoot the monitoring root
 * @param history        the history
 * @param network        the network
 */
public void report(@Nonnull final NotebookOutput log, @Nonnull final MonitoredObject monitoringRoot, @Nonnull final List<Step> history, @Nonnull final Layer network) {
    if (!history.isEmpty()) {
        log.code(() -> {
            @Nonnull final PlotCanvas plot = ScatterPlot.plot(history.stream().map(step -> new double[] { step.iteration, Math.log10(step.point.getMean()) }).toArray(i -> new double[i][]));
            plot.setTitle("Convergence Plot");
            plot.setAxisLabels("Iteration", "log10(Fitness)");
            plot.setSize(600, 400);
            return plot;
        });
    }
    @Nonnull final String modelName = "model" + modelNo++ + ".json";
    log.p("Saved model as " + log.file(network.getJson().toString(), modelName, modelName));
    log.h1("Metrics");
    log.code(() -> {
        try {
            @Nonnull final ByteArrayOutputStream out = new ByteArrayOutputStream();
            JsonUtil.writeJson(out, monitoringRoot.getMetrics());
            return out.toString();
        } catch (@Nonnull final IOException e) {
            throw new RuntimeException(e);
        }
    });
}

Example 30 with Layer

use of com.simiacryptus.mindseye.lang.Layer in project MindsEye by SimiaCryptus.

the class LBFGS method orient.

@Override
public SimpleLineSearchCursor orient(final Trainable subject, @Nonnull final PointSample measurement, @Nonnull final TrainingMonitor monitor) {
    // if (getClass().desiredAssertionStatus()) {
    // double verify = subject.measure(monitor).getMean();
    // double input = measurement.getMean();
    // boolean isDifferent = Math.abs(verify - input) > 1e-2;
    // if (isDifferent) throw new AssertionError(String.format("Invalid input point: %s != %s", verify, input));
    // monitor.log(String.format("Verified input point: %s == %s", verify, input));
    // }
    addToHistory(measurement, monitor);
    @Nonnull final List<PointSample> history = Arrays.asList(this.history.toArray(new PointSample[] {}));
    @Nullable final DeltaSet<Layer> result = lbfgs(measurement, monitor, history);
    SimpleLineSearchCursor returnValue;
    if (null == result) {
        @Nonnull DeltaSet<Layer> scale = measurement.delta.scale(-1);
        returnValue = cursor(subject, measurement, "GD", scale);
        scale.freeRef();
    } else {
        returnValue = cursor(subject, measurement, "LBFGS", result);
        result.freeRef();
    }
    while (this.history.size() > (null == result ? minHistory : maxHistory)) {
        @Nullable final PointSample remove = this.history.pollFirst();
        if (verbose) {
            monitor.log(String.format("Removed measurement %s to history. Total: %s", Long.toHexString(System.identityHashCode(remove)), history.size()));
        }
        remove.freeRef();
    }
    return returnValue;
}