predication | Create complex predicates from data | Data Visualization library

One possible error is in the calculation of delta:

You could try adding the transform-es2015-modules-commonjs plugin to your babel config file for testing only. Here is an example config file which tells babel to transpile modules only when in a testing environment. You can put it underneath your presets:

Either use TF in the browser via webpack and operate DOM as usual

or if you want a server-side fun use cheerio library.

You should do multiple steps :

1- scale the numeric features and one-hot encode the categorical ones (you can also encode your categorical variables with their number of appearances = replace each value with it's count)

2- study correlation between your target and other variables

3- Use different plots to know your data better

4- Use variable selection methods while modeling

EXISTS clause always performs faster than IN clause. You must give that a try -

I'm not sure of what you mean by "Frama-C wasn't too happy about the syntax". Your predicate looks syntactically correct to me, and to my Frama-C version as well.

Semantically, though, there is indeed an issue: you shouldn't use the implication (==>) but a conjunction (&&) under the existential quantifier. Otherwise, any size<=k<=l would be a witness satisfying the formula.

More generally, you almost always use quantifications like \forall x, P(x) ==> Q(x) and \exists x, P(x) && Q(x). Indeed, the former reads "for any x, if P(x) holds, then Q(x) holds, while the latter is "I want to find an x verifying both P(x) and Q(x). If you replace the conjunction by an implication, you're asking for an x such that if P(x) holds then so is Q(x), which can be achieved (in classical logic at least) by finding an x for which P(x) does not hold.

That said, automated provers often have difficulties with existential quantifiers (because they basically have to exhibit some witness for the formula), and according to your informal spec, there there is a pair (k,l) which is obvious: 0 and size-1. Of course, you need for that to add to your predicate the condition that size>=2, but you need it anyway, else you would face the same issue that both predicates are true for a single-element array. By the way, you probably also need to add size>=1 in the is_sortedInc predicate as well. Otherwise, the predicate will be true for size==0 (a universal quantification over an empty set of values is always true), but your function returns 0 in that case, so that the corresponding ensures does not hold.

I haven't checked your loop invariants in detail, but they look quite reasonable.

UPDATE Based on your comment below, your new version of the predicates still makes some confusion on the use of connectors and quantifiers:

• the conditions on size itself should be outside of any quantifier.
• In isSortedDec, you should have an implication under the forall and a conjunction under the exists, which itself should not be under the forall.

To summarize, the predicates should rather look like

indeed for detection you need to finetune the network, since you are using SSD, there are already some sources out there:

• https://gluon-cv.mxnet.io/build/examples_detection/finetune_detection.html (This one specifically for an SSD Model, uses mxnet but you can use the same with TF)
• You can watch a very nice finetunning intro here
• This repo has a nice fine tunning option enabled as long as you write your dataloader, check it out here

In general your error can be attributed to many factors, the learning rate you are using, the characteristics of the images themselves (are they normalized?) If the ssd network you are using was trained with normalized data and you don't normalize to retrain then you'll get stuck while learning. Also what learning rate are they using?

From the model zoo I can see that for SSD there are models trained on COCO

And models trained on Open Images:

If for example you are using ssd_inception_v2_coco, there is a truncated_normal_initializer in the input layers, so take that into consideration, also make sure the input sizes are the same that the ones you provide to the model.

You can get very good detections even with little data if you also include many augmentations and take into account the rest of the things I mentioned, more details on your code would help to see where the problem lies.

It tells you the size of the original image is 0, corresponding to when you read from img_array = cv2.imread(filepath, cv2.IMREAD_GRAYSCALE). Check if you do have that picture Dog.jpg loaded.

One option here would be to use MySQL's SELECT ... INTO OUTFILE capability (but see below for notes):